ICML 2018 论文列表

Phase retrieval algorithms have become an important component in many modern computational imaging systems. For instance, in the context of ptychography and speckle correlation imaging, they enable imaging past the diffraction limit and through scattering media, respectively. Unfortunately, traditional phase retrieval algorithms struggle in the presence of noise. Progress has been made recently on developing more robust algorithms using signal priors, but at the expense of limiting the range of supported measurement models (e.g., to Gaussian or coded diffraction patterns). In this work we leverage the regularization-by-denoising framework and a convolutional neural network denoiser to create prDeep, a new phase retrieval algorithm that is both robust and broadly applicable. We test and validate prDeep in simulation to demonstrate that it is robust to noise and can handle a variety of system models.

Deep generative models have recently yielded encouraging results in producing subjectively realistic samples of complex data. Far less attention has been paid to making these generative models interpretable. In many scenarios, ranging from scientific applications to finance, the observed variables have a natural grouping. It is often of interest to understand systems of interaction amongst these groups, and latent factor models (LFMs) are an attractive approach. However, traditional LFMs are limited by assuming a linear correlation structure. We present an output interpretable VAE (oi-VAE) for grouped data that models complex, nonlinear latent-to-observed relationships. We combine a structured VAE comprised of group-specific generators with a sparsity-inducing prior. We demonstrate that oi-VAE yields meaningful notions of interpretability in the analysis of motion capture and MEG data. We further show that in these situations, the regularization inherent to oi-VAE can actually lead to improved generalization and learned generative processes.

While it’s always possible to compute a variational approximation to a posterior distribution, it can be difficult to discover problems with this approximation. We propose two diagnostic algorithms to alleviate this problem. The Pareto-smoothed importance sampling (PSIS) diagnostic gives a goodness of fit measurement for joint distributions, while simultaneously improving the error in the estimate. The variational simulation-based calibration (VSBC) assesses the average performance of point estimates.

Recent work has shown local convergence of GAN training for absolutely continuous data and generator distributions. In this paper, we show that the requirement of absolute continuity is necessary: we describe a simple yet prototypical counterexample showing that in the more realistic case of distributions that are not absolutely continuous, unregularized GAN training is not always convergent. Furthermore, we discuss regularization strategies that were recently proposed to stabilize GAN training. Our analysis shows that GAN training with instance noise or zero-centered gradient penalties converges. On the other hand, we show that Wasserstein-GANs and WGAN-GP with a finite number of discriminator updates per generator update do not always converge to the equilibrium point. We discuss these results, leading us to a new explanation for the stability problems of GAN training. Based on our analysis, we extend our convergence results to more general GANs and prove local convergence for simplified gradient penalties even if the generator and data distributions lie on lower dimensional manifolds. We find these penalties to work well in practice and use them to learn high-resolution generative image models for a variety of datasets with little hyperparameter tuning.

The large memory requirements of deep neural networks limit their deployment and adoption on many devices. Model compression methods effectively reduce the memory requirements of these models, usually through applying transformations such as weight pruning or quantization. In this paper, we present a novel scheme for lossy weight encoding co-designed with weight simplification techniques. The encoding is based on the Bloomier filter, a probabilistic data structure that can save space at the cost of introducing random errors. Leveraging the ability of neural networks to tolerate these imperfections and by re-training around the errors, the proposed technique, named Weightless, can compress weights by up to 496x without loss of model accuracy. This results in up to a 1.51x improvement over the state-of-the-art.

Submodular functions are a broad class of set functions that naturally arise in many machine learning applications. Due to their combinatorial structures, there has been a myriad of algorithms for maximizing such functions under various constraints. Unfortunately, once a function deviates from submodularity (even slightly), the known algorithms may perform arbitrarily poorly. Amending this issue, by obtaining approximation results for functions obeying properties that generalize submodularity, has been the focus of several recent works. One such class, known as weakly submodular functions, has received a lot of recent attention from the machine learning community due to its strong connections to restricted strong convexity and sparse reconstruction. In this paper, we prove that a randomized version of the greedy algorithm achieves an approximation ratio of $(1 + 1/\gamma )^{-2}$ for weakly submodular maximization subject to a general matroid constraint, where $\gamma$ is a parameter measuring the distance from submodularity. To the best of our knowledge, this is the first algorithm with a non-trivial approximation guarantee for this constrained optimization problem. Moreover, our experimental results show that our proposed algorithm performs well in a variety of real-world problems, including regression, video summarization, splice site detection, and black-box interpretation.

We study revenue optimization learning algorithms for repeated posted-price auctions where a seller interacts with a single strategic buyer that holds a fixed private valuation for a good and seeks to maximize his cumulative discounted surplus. We propose a novel algorithm that never decreases offered prices and has a tight strategic regret bound of $\Theta(\log\log T)$. This result closes the open research question on the existence of a no-regret horizon-independent weakly consistent pricing. We also show that the property of non-decreasing prices is nearly necessary for a weakly consistent algorithm to be a no-regret one.

We present a new approach and a novel architecture, termed WSNet, for learning compact and efficient deep neural networks. Existing approaches conventionally learn full model parameters independently and then compress them via ad hoc processing such as model pruning or filter factorization. Alternatively, WSNet proposes learning model parameters by sampling from a compact set of learnable parameters, which naturally enforces parameter sharing throughout the learning process. We demonstrate that such a novel weight sampling approach (and induced WSNet) promotes both weights and computation sharing favorably. By employing this method, we can more efficiently learn much smaller networks with competitive performance compared to baseline networks with equal numbers of convolution filters. Specifically, we consider learning compact and efficient 1D convolutional neural networks for audio classification. Extensive experiments on multiple audio classification datasets verify the effectiveness of WSNet. Combined with weight quantization, the resulted models are up to 180x smaller and theoretically up to 16x faster than the well-established baselines, without noticeable performance drop.

Learning sparse linear models with two-way interactions is desirable in many application domains such as genomics. $\ell_1$-regularised linear models are popular to estimate sparse models, yet standard implementations fail to address specifically the quadratic explosion of candidate two-way interactions in high dimensions, and typically do not scale to genetic data with hundreds of thousands of features. Here we present WHInter, a working set algorithm to solve large $\ell_1$-regularised problems with two-way interactions for binary design matrices. The novelty of WHInter stems from a new bound to efficiently identify working sets while avoiding to scan all features, and on fast computations inspired from solutions to the maximum inner product search problem. We apply WHInter to simulated and real genetic data and show that it is more scalable and two orders of magnitude faster than the state of the art.

While deep reinforcement learning (deep RL) agents are effective at maximizing rewards, it is often unclear what strategies they use to do so. In this paper, we take a step toward explaining deep RL agents through a case study using Atari 2600 environments. In particular, we focus on using saliency maps to understand how an agent learns and executes a policy. We introduce a method for generating useful saliency maps and use it to show 1) what strong agents attend to, 2) whether agents are making decisions for the right or wrong reasons, and 3) how agents evolve during learning. We also test our method on non-expert human subjects and find that it improves their ability to reason about these agents. Overall, our results show that saliency information can provide significant insight into an RL agent’s decisions and learning behavior.

Video prediction aims to generate realistic future frames by learning dynamic visual patterns. One fundamental challenge is to deal with future uncertainty: How should a model behave when there are multiple correct, equally probable future? We propose an Appearance-Motion Conditional GAN to address this challenge. We provide appearance and motion information as conditions that specify how the future may look like, reducing the level of uncertainty. Our model consists of a generator, two discriminators taking charge of appearance and motion pathways, and a perceptual ranking module that encourages videos of similar conditions to look similar. To train our model, we develop a novel conditioning scheme that consists of different combinations of appearance and motion conditions. We evaluate our model using facial expression and human action datasets and report favorable results compared to existing methods.

Traditional methods for the discovery of latent network structures are limited in two ways: they either assume that all the signal comes from the network (i.e. there is no source of signal outside the network) or they place constraints on the network parameters to ensure model or algorithmic stability. We address these limitations by proposing a model that incorporates a Gaussian process prior on a network-independent component and formally proving that we get algorithmic stability for free while providing a novel perspective on model stability as well as robustness results and precise intervals for key inference parameters. We show that, on three applications, our approach outperforms previous methods consistently.

Recent advances on the scalability and flexibility of variational inference have made it successful at unravelling hidden patterns in complex data. In this work we propose a new variational bound formulation, yielding an estimator that extends beyond the conventional variational bound. It naturally subsumes the importance-weighted and Renyi bounds as special cases, and it is provably sharper than these counterparts. We also present an improved estimator for variational learning, and advocate a novel high signal-to-variance ratio update rule for the variational parameters. We discuss model-selection issues associated with existing evidence-lower-bound-based variational inference procedures, and show how to leverage the flexibility of our new formulation to address them. Empirical evidence is provided to validate our claims.

Dropout, a stochastic regularisation technique for training of neural networks, has recently been reinterpreted as a specific type of approximate inference algorithm for Bayesian neural networks. The main contribution of the reinterpretation is in providing a theoretical framework useful for analysing and extending the algorithm. We show that the proposed framework suffers from several issues; from undefined or pathological behaviour of the true posterior related to use of improper priors, to an ill-defined variational objective due to singularity of the approximating distribution relative to the true posterior. Our analysis of the improper log uniform prior used in variational Gaussian dropout suggests the pathologies are generally irredeemable, and that the algorithm still works only because the variational formulation annuls some of the pathologies. To address the singularity issue, we proffer Quasi-KL (QKL) divergence, a new approximate inference objective for approximation of high-dimensional distributions. We show that motivations for variational Bernoulli dropout based on discretisation and noise have QKL as a limit. Properties of QKL are studied both theoretically and on a simple practical example which shows that the QKL-optimal approximation of a full rank Gaussian with a degenerate one naturally leads to the Principal Component Analysis solution.

Off-policy learning, the task of evaluating and improving policies using historic data collected from a logging policy, is important because on-policy evaluation is usually expensive and has adverse impacts. One of the major challenge of off-policy learning is to derive counterfactual estimators that also has low variance and thus low generalization error. In this work, inspired by learning bounds for importance sampling problems, we present a new counterfactual learning principle for off-policy learning with bandit feedbacks. Our method regularizes the generalization error by minimizing the distribution divergence between the logging policy and the new policy, and removes the need for iterating through all training samples to compute sample variance regularization in prior work. With neural network policies, our end-to-end training algorithms using variational divergence minimization showed significant improvement over conventional baseline algorithms and is also consistent with our theoretical results.

We propose a variable selection method for high dimensional regression models, which allows for complex, nonlinear, and high-order interactions among variables. The proposed method approximates this complex system using a penalized neural network and selects explanatory variables by measuring their utility in explaining the variance of the response variable. This measurement is based on a novel statistic called Drop-Out-One Loss. The proposed method also allows (overlapping) group variable selection. We prove that the proposed method can select relevant variables and exclude irrelevant variables with probability one as the sample size goes to infinity, which is referred to as the Oracle Property. Experimental results on simulated and real world datasets show the efficiency of our method in terms of variable selection and prediction accuracy.

In this paper we propose Reward Machines {—} a type of finite state machine that supports the specification of reward functions while exposing reward function structure to the learner and supporting decomposition. We then present Q-Learning for Reward Machines (QRM), an algorithm which appropriately decomposes the reward machine and uses off-policy q-learning to simultaneously learn subpolicies for the different components. QRM is guaranteed to converge to an optimal policy in the tabular case, in contrast to Hierarchical Reinforcement Learning methods which might converge to suboptimal policies. We demonstrate this behavior experimentally in two discrete domains. We also show how function approximation methods like neural networks can be incorporated into QRM, and that doing so can find better policies more quickly than hierarchical methods in a domain with a continuous state space.

Understanding the representational power of Restricted Boltzmann Machines (RBMs) with multiple layers is an ill-understood problem and is an area of active research. Motivated from the approach of

A key challenge in complex visuomotor control is learning abstract representations that are effective for specifying goals, planning, and generalization. To this end, we introduce universal planning networks (UPN). UPNs embed differentiable planning within a goal-directed policy. This planning computation unrolls a forward model in a latent space and infers an optimal action plan through gradient descent trajectory optimization. The plan-by-gradient-descent process and its underlying representations are learned end-to-end to directly optimize a supervised imitation learning objective. We find that the representations learned are not only effective for goal-directed visual imitation via gradient-based trajectory optimization, but can also provide a metric for specifying goals using images. The learned representations can be leveraged to specify distance-based rewards to reach new target states for model-free reinforcement learning, resulting in substantially more effective learning when solving new tasks described via image based goals. We were able to achieve successful transfer of visuomotor planning strategies across robots with significantly different morphologies and actuation capabilities. Visit https://sites.google. com/view/upn-public/home for video highlights.