ICML 2016 论文列表

Neural network architectures with memory and attention mechanisms exhibit certain reason- ing capabilities required for question answering. One such architecture, the dynamic memory net- work (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.

We introduce the Dynamic Capacity Network (DCN), a neural network that can adaptively assign its capacity across different portions of the input data. This is achieved by combining modules of two types: low-capacity sub-networks and high-capacity sub-networks. The low-capacity sub-networks are applied across most of the input, but also provide a guide to select a few portions of the input on which to apply the high-capacity sub-networks. The selection is made using a novel gradient-based attention mechanism, that efficiently identifies input regions for which the DCN’s output is most sensitive and to which we should devote more capacity. We focus our empirical evaluation on the Cluttered MNIST and SVHN image datasets. Our findings indicate that DCNs are able to drastically reduce the number of computations, compared to traditional convolutional neural networks, while maintaining similar or even better performance.

In recent years there have been many successes of using deep representations in reinforcement learning. Still, many of these applications use conventional architectures, such as convolutional networks, LSTMs, or auto-encoders. In this paper, we present a new neural network architecture for model-free reinforcement learning. Our dueling network represents two separate estimators: one for the state value function and one for the state-dependent action advantage function. The main benefit of this factoring is to generalize learning across actions without imposing any change to the underlying reinforcement learning algorithm. Our results show that this architecture leads to better policy evaluation in the presence of many similar-valued actions. Moreover, the dueling architecture enables our RL agent to outperform the state-of-the-art on the Atari 2600 domain.

Dropout is a popular stochastic regularization technique for deep neural networks that works by randomly dropping (i.e. zeroing) units from the network during training. This randomization process allows to implicitly train an ensemble of exponentially many networks sharing the same parametrization, which should be averaged at test time to deliver the final prediction. A typical workaround for this intractable averaging operation consists in scaling the layers undergoing dropout randomization. This simple rule called ’standard dropout’ is efficient, but might degrade the accuracy of the prediction. In this work we introduce a novel approach, coined ’dropout distillation’, that allows us to train a predictor in a way to better approximate the intractable, but preferable, averaging process, while keeping under control its computational efficiency. We are thus able to construct models that are as efficient as standard dropout, or even more efficient, while being more accurate. Experiments on standard benchmark datasets demonstrate the validity of our method, yielding consistent improvements over conventional dropout.

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.

We study the problem of off-policy value evaluation in reinforcement learning (RL), where one aims to estimate the value of a new policy based on data collected by a different policy. This problem is often a critical step when applying RL to real-world problems. Despite its importance, existing general methods either have uncontrolled bias or suffer high variance. In this work, we extend the doubly robust estimator for bandits to sequential decision-making problems, which gets the best of both worlds: it is guaranteed to be unbiased and can have a much lower variance than the popular importance sampling estimators. We demonstrate the estimator’s accuracy in several benchmark problems, and illustrate its use as a subroutine in safe policy improvement. We also provide theoretical results on the inherent hardness of the problem, and show that our estimator can match the lower bound in certain scenarios.

Nonparametric extension of tensor regression is proposed. Nonlinearity in a high-dimensional tensor space is broken into simple local functions by incorporating low-rank tensor decomposition. Compared to naive nonparametric approaches, our formulation considerably improves the convergence rate of estimation while maintaining consistency with the same function class under specific conditions. To estimate local functions, we develop a Bayesian estimator with the Gaussian process prior. Experimental results show its theoretical properties and high performance in terms of predicting a summary statistic of a real complex network.

Domain adaptation arises in supervised learning when the training (source domain) and test (target domain) data have different distributions. Let X and Y denote the features and target, respectively, previous work on domain adaptation considers the covariate shift situation where the distribution of the features P(X) changes across domains while the conditional distribution P(Y|X) stays the same. To reduce domain discrepancy, recent methods try to find invariant components \mathcalT(X) that have similar P(\mathcalT(X)) by explicitly minimizing a distribution discrepancy measure. However, it is not clear if P(Y|\mathcalT(X)) in different domains is also similar when P(Y|X) changes. Furthermore, transferable components do not necessarily have to be invariant. If the change in some components is identifiable, we can make use of such components for prediction in the target domain. In this paper, we focus on the case where P(X|Y) and P(Y) both change in a causal system in which Y is the cause for X. Under appropriate assumptions, we aim to extract conditional transferable components whose conditional distribution P(\mathcalT(X)|Y) is invariant after proper location-scale (LS) transformations, and identify how P(Y) changes between domains simultaneously. We provide theoretical analysis and empirical evaluation on both synthetic and real-world data to show the effectiveness of our method.

In learning latent variable models (LVMs), it is important to effectively capture infrequent patterns and shrink model size without sacrificing modeling power. Various studies have been done to “diversify” a LVM, which aim to learn a diverse set of latent components in LVMs. Most existing studies fall into a frequentist-style regularization framework, where the components are learned via point estimation. In this paper, we investigate how to “diversify” LVMs in the paradigm of Bayesian learning, which has advantages complementary to point estimation, such as alleviating overfitting via model averaging and quantifying uncertainty. We propose two approaches that have complementary advantages. One is to define diversity-promoting mutual angular priors which assign larger density to components with larger mutual angles based on Bayesian network and von Mises-Fisher distribution and use these priors to affect the posterior via Bayes rule. We develop two efficient approximate posterior inference algorithms based on variational inference and Markov chain Monte Carlo sampling. The other approach is to impose diversity-promoting regularization directly over the post-data distribution of components. These two methods are applied to the Bayesian mixture of experts model to encourage the “experts” to be diverse and experimental results demonstrate the effectiveness and efficiency of our methods.

We provide two distributed confidence ball algorithms for solving linear bandit problems in peer to peer networks with limited communication capabilities. For the first, we assume that all the peers are solving the same linear bandit problem, and prove that our algorithm achieves the optimal asymptotic regret rate of any centralised algorithm that can instantly communicate information between the peers. For the second, we assume that there are clusters of peers solving the same bandit problem within each cluster, and we prove that our algorithm discovers these clusters, while achieving the optimal asymptotic regret rate within each one. Through experiments on several real-world datasets, we demonstrate the performance of proposed algorithms compared to the state-of-the-art.

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.

The collection and analysis of user data drives improvements in the app and web ecosystems, but comes with risks to privacy. This paper examines discrete distribution estimation under local privacy, a setting wherein service providers can learn the distribution of a categorical statistic of interest without collecting the underlying data. We present new mechanisms, including hashed k-ary Randomized Response (KRR), that empirically meet or exceed the utility of existing mechanisms at all privacy levels. New theoretical results demonstrate the order-optimality of KRR and the existing RAPPOR mechanism at different privacy regimes.

First steps towards a mathematical theory of deep convolutional neural networks for feature extraction were made—for the continuous-time case—in Mallat, 2012, and Wiatowski and Bölcskei, 2015. This paper considers the discrete case, introduces new convolutional neural network architectures, and proposes a mathematical framework for their analysis. Specifically, we establish deformation and translation sensitivity results of local and global nature, and we investigate how certain structural properties of the input signal are reflected in the corresponding feature vectors. Our theory applies to general filters and general Lipschitz-continuous non-linearities and pooling operators. Experiments on handwritten digit classification and facial landmark detection—including feature importance evaluation—complement the theoretical findings.

We introduce an iterative normalization and clustering method for single-cell gene expression data. The emerging technology of single-cell RNA-seq gives access to gene expression measurements for thousands of cells, allowing discovery and characterization of cell types. However, the data is confounded by technical variation emanating from experimental errors and cell type-specific biases. Current approaches perform a global normalization prior to analyzing biological signals, which does not resolve missing data or variation dependent on latent cell types. Our model is formulated as a hierarchical Bayesian mixture model with cell-specific scalings that aid the iterative normalization and clustering of cells, teasing apart technical variation from biological signals. We demonstrate that this approach is superior to global normalization followed by clustering. We show identifiability and weak convergence guarantees of our method and present a scalable Gibbs inference algorithm. This method improves cluster inference in both synthetic and real single-cell data compared with previous methods, and allows easy interpretation and recovery of the underlying structure and cell types.

We present the first differentially private algorithms for reinforcement learning, which apply to the task of evaluating a fixed policy. We establish two approaches for achieving differential privacy, provide a theoretical analysis of the privacy and utility of the two algorithms, and show promising results on simple empirical examples.

Hypothesis testing is a useful statistical tool in determining whether a given model should be rejected based on a sample from the population. Sample data may contain sensitive information about individuals, such as medical information. Thus it is important to design statistical tests that guarantee the privacy of subjects in the data. In this work, we study hypothesis testing subject to differential privacy, specifically chi-squared tests for goodness of fit for multinomial data and independence between two categorical variables.

We study the problem of supervised learning for both binary and multiclass classification from a unified geometric perspective. In particular, we propose a geometric regularization technique to find the submanifold corresponding to an estimator of the class probability P(y|\vec x). The regularization term measures the volume of this submanifold, based on the intuition that overfitting produces rapid local oscillations and hence large volume of the estimator. This technique can be applied to regularize any classification function that satisfies two requirements: firstly, an estimator of the class probability can be obtained; secondly, first and second derivatives of the class probability estimator can be calculated. In experiments, we apply our regularization technique to standard loss functions for classification, our RBF-based implementation compares favorably to widely used regularization methods for both binary and multiclass classification.

Sparse matrix factorization is a popular tool to obtain interpretable data decompositions, which are also effective to perform data completion or denoising. Its applicability to large datasets has been addressed with online and randomized methods, that reduce the complexity in one of the matrix dimension, but not in both of them. In this paper, we tackle very large matrices in both dimensions. We propose a new factorization method that scales gracefully to terabyte-scale datasets. Those could not be processed by previous algorithms in a reasonable amount of time. We demonstrate the efficiency of our approach on massive functional Magnetic Resonance Imaging (fMRI) data, and on matrix completion problems for recommender systems, where we obtain significant speed-ups compared to state-of-the art coordinate descent methods.

In this paper, we attack the anomaly detection problem by directly modeling the data distribution with deep architectures. We hence propose deep structured energy based models (DSEBMs), where the energy function is the output of a deterministic deep neural network with structure. We develop novel model architectures to integrate EBMs with different types of data such as static data, sequential data, and spatial data, and apply appropriate model architectures to adapt to the data structure. Our training algorithm is built upon the recent development of score matching (Hyvarinen, 2005), which connects an EBM with a regularized autoencoder, eliminating the need for complicated sampling method. Statistically sound decision criterion can be derived for anomaly detection purpose from the perspective of the energy landscape of the data distribution. We investigate two decision criteria for performing anomaly detection: the energy score and the reconstruction error. Extensive empirical studies on benchmark anomaly detection tasks demonstrate that our proposed model consistently matches or outperforms all the competing methods.