ICML 2017 论文列表

Curating labeled training data has become the primary bottleneck in machine learning. Recent frameworks address this bottleneck with generative models to synthesize labels at scale from weak supervision sources. The generative model’s dependency structure directly affects the quality of the estimated labels, but selecting a structure automatically without any labeled data is a distinct challenge. We propose a structure estimation method that maximizes the l1-regularized marginal pseudolikelihood of the observed data. Our analysis shows that the amount of unlabeled data required to identify the true structure scales sublinearly in the number of possible dependencies for a broad class of models. Simulations show that our method is 100x faster than a maximum likelihood approach and selects 1/4 as many extraneous dependencies. We also show that our method provides an average of 1.5 F1 points of improvement over existing, user-developed information extraction applications on real-world data such as PubMed journal abstracts.

The POMDP is a powerful framework for reasoning under outcome and information uncertainty, but constructing an accurate POMDP model is difficult. Bayes-Adaptive Partially Observable Markov Decision Processes (BA-POMDPs) extend POMDPs to allow the model to be learned during execution. BA-POMDPs are a Bayesian RL approach that, in principle, allows for an optimal trade-off between exploitation and exploration. Unfortunately, BA-POMDPs are currently impractical to solve for any non-trivial domain. In this paper, we extend the Monte-Carlo Tree Search method POMCP to BA-POMDPs and show that the resulting method, which we call BA-POMCP, is able to tackle problems that previous solution methods have been unable to solve. Additionally, we introduce several techniques that exploit the BA-POMDP structure to improve the efficiency of BA-POMCP along with proof of their convergence.

Critically ill patients in regular wards are vulnerable to unanticipated adverse events which require prompt transfer to the intensive care unit (ICU). To allow for accurate prognosis of deteriorating patients, we develop a novel continuous-time probabilistic model for a monitored patient’s temporal sequence of physiological data. Our model captures “informatively sampled” patient episodes: the clinicians’ decisions on when to observe a hospitalized patient’s vital signs and lab tests over time are represented by a marked Hawkes process, with intensity parameters that are modulated by the patient’s latent clinical states, and with observable physiological data (mark process) modeled as a switching multi-task Gaussian process. In addition, our model captures “informatively censored” patient episodes by representing the patient’s latent clinical states as an absorbing semi-Markov jump process. The model parameters are learned from offline patient episodes in the electronic health records via an EM-based algorithm. Experiments conducted on a cohort of patients admitted to a major medical center over a 3-year period show that risk prognosis based on our model significantly outperforms the currently deployed medical risk scores and other baseline machine learning algorithms.

This paper introduces a novel approach to texture synthesis based on generative adversarial networks (GAN) (Goodfellow et al., 2014), and call this technique Periodic Spatial GAN (PSGAN). The PSGAN has several novel abilities which surpass the current state of the art in texture synthesis. First, we can learn multiple textures, periodic or non-periodic, from datasets of one or more complex large images. Second, we show that the image generation with PSGANs has properties of a texture manifold: we can smoothly interpolate between samples in the structured noise space and generate novel samples, which lie perceptually between the textures of the original dataset. We make multiple experiments which show that PSGANs can flexibly handle diverse texture and image data sources, and the method is highly scalable and can generate output images of arbitrary large size.

A data-driven identification of dynamical systems requiring only minimal prior knowledge is promising whenever no analytically derived model structure is available, e.g., from first principles in physics. However, meta-knowledge on the system’s behavior is often given and should be exploited: Stability as fundamental property is essential when the model is used for controller design or movement generation. Therefore, this paper proposes a framework for learning stable stochastic systems from data. We focus on identifying a state-dependent coefficient form of the nonlinear stochastic model which is globally asymptotically stable according to probabilistic Lyapunov methods. We compare our approach to other state of the art methods on real-world datasets in terms of flexibility and stability.

We focus on predicting sleep stages from radio measurements without any attached sensors on subjects. We introduce a new predictive model that combines convolutional and recurrent neural networks to extract sleep-specific subject-invariant features from RF signals and capture the temporal progression of sleep. A key innovation underlying our approach is a modified adversarial training regime that discards extraneous information specific to individuals or measurement conditions, while retaining all information relevant to the predictive task. We analyze our game theoretic setup and empirically demonstrate that our model achieves significant improvements over state-of-the-art solutions.

The large model capacity of latent space models (LSMs) enables them to achieve great performance on various applications, but meanwhile renders LSMs to be prone to overfitting. Several recent studies investigate a new type of regularization approach, which encourages components in LSMs to be diverse, for the sake of alleviating overfitting. While they have shown promising empirical effectiveness, in theory why larger “diversity” results in less overfitting is still unclear. To bridge this gap, we propose a new diversity-promoting approach that is both theoretically analyzable and empirically effective. Specifically, we use near-orthogonality to characterize “diversity” and impose angular constraints (ACs) on the components of LSMs to promote diversity. A generalization error analysis shows that larger diversity results in smaller estimation error and larger approximation error. An efficient ADMM algorithm is developed to solve the constrained LSM problems. Experiments demonstrate that ACs improve generalization performance of LSMs and outperform other diversity-promoting approaches.

Infinite Layer Networks (ILN) have 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.

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

Deep neural networks have been shown to be very successful at learning feature hierarchies in supervised learning tasks. Generative models, on the other hand, have benefited less from hierarchical models with multiple layers of latent variables. In this paper, we prove that hierarchical latent variable models do not take advantage of the hierarchical structure when trained with existing variational methods, and provide some limitations on the kind of features existing models can learn. Finally we propose an alternative architecture that do not suffer from these limitations. Our model is able to learn highly interpretable and disentangled hierarchical features on several natural image datasets with no task specific regularization or prior knowledge.

Many real-world applications require robust algorithms to learn point process models based on a type of incomplete data — the so-called short doubly-censored (SDC) event sequences. In this paper, we study this critical problem of quantitative asynchronous event sequence analysis under the framework of Hawkes processes by leveraging the general idea of data synthesis. In particular, given SDC event sequences observed in a variety of time intervals, we propose a sampling-stitching data synthesis method — sampling predecessor and successor for each SDC event sequence from potential candidates and stitching them together to synthesize long training sequences. The rationality and the feasibility of our method are discussed in terms of arguments based on likelihood. Experiments on both synthetic and real-world data demonstrate that the proposed data synthesis method improves learning results indeed for both time-invariant and time-varying Hawkes processes.

Training deep neural networks is a highly nontrivial task, involving carefully selecting appropriate training algorithms, scheduling step sizes and tuning other hyperparameters. Trying different combinations can be quite labor-intensive and time consuming. Recently, researchers have tried to use deep learning algorithms to exploit the landscape of the loss function of the training problem of interest, and learn how to optimize over it in an automatic way. In this paper, we propose a new learning-to-learn model and some useful and practical tricks. Our optimizer outperforms generic, hand-crafted optimization algorithms and state-of-the-art learning-to-learn optimizers by DeepMind in many tasks. We demonstrate the effectiveness of our algorithms on a number of tasks, including deep MLPs, CNNs, and simple LSTMs.

Learning discrete representations of data is a central machine learning task because of the compactness of the representations and ease of interpretation. The task includes clustering and hash learning as special cases. Deep neural networks are promising to be used because they can model the non-linearity of data and scale to large datasets. However, their model complexity is huge, and therefore, we need to carefully regularize the networks in order to learn useful representations that exhibit intended invariance for applications of interest. To this end, we propose a method called Information Maximizing Self-Augmented Training (IMSAT). In IMSAT, we use data augmentation to impose the invariance on discrete representations. More specifically, we encourage the predicted representations of augmented data points to be close to those of the original data points in an end-to-end fashion. At the same time, we maximize the information-theoretic dependency between data and their predicted discrete representations. Extensive experiments on benchmark datasets show that IMSAT produces state-of-the-art results for both clustering and unsupervised hash learning.

Determinantal Point Processes (DPPs) are a family of probabilistic models that have a repulsive behavior, and lend themselves naturally to many tasks in machine learning where returning a diverse set of objects is important. While there are fast algorithms for sampling, marginalization and conditioning, much less is known about learning the parameters of a DPP. Our contribution is twofold: (i) we establish the optimal sample complexity achievable in this problem and show that it is governed by a natural parameter, which we call the cycle sparsity; (ii) we propose a provably fast combinatorial algorithm that implements the method of moments efficiently and achieves optimal sample complexity. Finally, we give experimental results that confirm our theoretical findings.

Deep latent Gaussian models are powerful and popular probabilistic models of high-dimensional data. These models are almost always fit using variational expectation-maximization, an approximation to true maximum-marginal-likelihood estimation. In this paper, we propose a different approach: rather than use a variational approximation (which produces biased gradient signals), we use Markov chain Monte Carlo (MCMC, which allows us to trade bias for computation). We find that our MCMC-based approach has several advantages: it yields higher held-out likelihoods, produces sharper images, and does not suffer from the variational overpruning effect. MCMC’s additional computational overhead proves to be significant, but not prohibitive.

Whitened Neural Network (WNN) is a recent advanced deep architecture, which improves convergence and generalization of canonical neural networks by whitening their internal hidden representation. However, the whitening transformation increases computation time. Unlike WNN that reduced runtime by performing whitening every thousand iterations, which degenerates convergence due to the ill conditioning, we present generalized WNN (GWNN), which has three appealing properties. First, GWNN is able to learn compact representation to reduce computations. Second, it enables whitening transformation to be performed in a short period, preserving good conditioning. Third, we propose a data-independent estimation of the covariance matrix to further improve computational efficiency. Extensive experiments on various datasets demonstrate the benefits of GWNN.

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 network, 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.

We introduce a model that learns active learning algorithms via metalearning. For a distribution of related tasks, our model jointly learns: a data representation, an item selection heuristic, and a prediction function. Our model uses the item selection heuristic to construct a labeled support set for training the prediction function. Using the Omniglot and MovieLens datasets, we test our model in synthetic and practical settings.

Learning to learn has emerged as an important direction for achieving artificial intelligence. Two of the primary barriers to its adoption are an inability to scale to larger problems and a limited ability to generalize to new tasks. We introduce a learned gradient descent optimizer that generalizes well to new tasks, and which has significantly reduced memory and computation overhead. We achieve this by introducing a novel hierarchical RNN architecture, with minimal per-parameter overhead, augmented with additional architectural features that mirror the known structure of optimization tasks. We also develop a meta-training ensemble of small, diverse, optimization tasks capturing common properties of loss landscapes. The optimizer learns to outperform RMSProp/ADAM on problems in this corpus. More importantly, it performs comparably or better when applied to small convolutional neural networks, despite seeing no neural networks in its meta-training set. Finally, it generalizes to train Inception V3 and ResNet V2 architectures on the ImageNet dataset for thousands of steps, optimization problems that are of a vastly different scale than those it was trained on.