AISTATS 2018 Papers

Advances in recovering low-rank matrices from noisy observations have led to tractable algorithms for clustering from general pairwise labels with provable performance guarantees. Based on convex relaxation, it has been shown that the ground truth clusters can be recovered with high probability under a generalized stochastic block model by solving a semidefinite program. Although tractable, the algorithm is typically too slow for sufficiently large problems in practice. Inspired by recent advances in non-convex approaches to low-rank recovery problems, we propose an algorithm based on projected gradient descent that enjoys similar provable guarantees as the convex counterpart, but can be orders of magnitude faster. Our theoretical results are further supported by encouraging empirical results.

A nonparametric family of conditional distributions is introduced, which generalizes conditional exponential families using functional parameters in a suitable RKHS. An algorithm is provided for learning the generalized natural parameter, and consistency of the estimator is established in the well specified case. In experiments, the new method generally outperforms a competing approach with consistency guarantees, and is competitive with a deep conditional density model on datasets that exhibit abrupt transitions and heteroscedasticity.

In high-dimensional prediction problems, where the number of features may greatly exceed the number of training instances, fully Bayesian approach with a sparsifying prior is known to produce good results but is computationally challenging. To alleviate this computational burden, we propose to use a preprocessing step where we first apply a dimension reduction to the original data to reduce the number of features to something that is computationally conveniently handled by Bayesian methods. To do this, we propose a new dimension reduction technique, called iterative supervised principal components (ISPCs), which combines variable screening and dimension reduction and can be considered as an extension to the existing technique of supervised principal components (SPCs). Our empirical evaluations confirm that, although not foolproof, the proposed approach provides very good results on several microarray benchmark datasets with very affordable computation time, and it can also be very useful for visualizing high-dimensional data.

Given a dataset and an existing clustering as input, alternative clustering aims to find an alternative partition. One of the state-of-the-art approaches is Kernel Dimension Alternative Clustering (KDAC). We propose a novel Iterative Spectral Method (ISM) that greatly improves the scalability of KDAC. Our algorithm is intuitive, relies on easily implementable spectral decompositions, and comes with theoretical guarantees. Its computation time improves upon existing implementations of KDAC by as much as 5 orders of magnitude.

To compare learning algorithms that differ by the adopted statistical paradigm, model class, or search heuristic, it is common to evaluate the performance on training data of varying size. Measuring the performance is straightforward if the data are generated from a known model, the ground truth. However, when the study concerns real-world data, the current methodology is limited to estimating predictive performance, typically by cross-validation. This work introduces a method to compare algorithms’ ability to learn the model structure, assuming no ground truth is given. The idea is to identify a partial structure on which the algorithms agree, and measure the performance in relation to that structure on subsamples of the data. The method is instantiated to structure learning in Bayesian networks, measuring the performance by the structural Hamming distance. It is tested using benchmark ground truth networks and algorithms that maximize various scoring functions. The results show that the method can produce evaluation outcomes that are close to those one would obtain if the ground truth was available.

Computing accurate estimates of the Fourier transform of analog signals from discrete data points is important in many fields of science and engineering. The conventional approach of performing the discrete Fourier transform of the data implicitly assumes periodicity and bandlimitedness of the signal. In this paper, we use Gaussian process regression to estimate the Fourier transform (or any other integral transform) without making these assumptions. This is possible because the posterior expectation of Gaussian process regression maps a finite set of samples to a function defined on the whole real line, expressed as a linear combination of covariance functions. We estimate the covariance function from the data using an appropriately designed gradient ascent method that constrains the solution to a linear combination of tractable kernel functions. This procedure results in a posterior expectation of the analog signal whose Fourier transform can be obtained analytically by exploiting linearity. Our simulations show that the new method leads to sharper and more precise estimation of the spectral density both in noise-free and noise-corrupted signals. We further validate the method in two real-world applications: the analysis of the yearly fluctuation in atmospheric CO2 level and the analysis of the spectral content of brain signals.

We consider the statistical problem of recovering a hidden "ground truth" binary labeling for the vertices of a graph up to low Hamming error from noisy edge and vertex measurements. We present new algorithms and a sharp finite-sample analysis for this problem on trees and sparse graphs with poor expansion properties such as hypergrids and ring lattices. Our method generalizes and improves over that of Globerson et al. (2015), who introduced the problem for two-dimensional grid lattices. For trees we provide a simple, efficient, algorithm that infers the ground truth with optimal Hamming error has optimal sample complexity and implies recovery results for all connected graphs. Here, the presence of side information is critical to obtain a non-trivial recovery rate. We then show how to adapt this algorithm to tree decompositions of edge-subgraphs of certain graph families such as lattices, resulting in optimal recovery error rates that can be obtained efficiently The thrust of our analysis is to 1) use the tree decomposition along with edge measurements to produce a small class of viable vertex labelings and 2) apply an analysis influenced by statistical learning theory to show that we can infer the ground truth from this class using vertex measurements. We show the power of our method in several examples including hypergrids, ring lattices, and the Newman-Watts model for small world graphs. For two-dimensional grids, our results improve over Globerson et al. (2015) by obtaining optimal recovery in the constant-height regime.

Sparse regularization such as l1 regularization is a quite powerful and widely used strategy for high dimensional learning problems. The effectiveness of sparse regularization has been supported practically and theoretically by several studies. However, one of the biggest issues in sparse regularization is that its performance is quite sensitive to correlations between features. Ordinary l1 regularization can select variables correlated with each other, which results in deterioration of not only its generalization error but also interpretability. In this paper, we pro- pose a new regularization method, “Independently Interpretable Lasso” (IILasso). Our proposed regularizer suppresses selecting correlated variables, and thus each active variable independently affects the objective variable in the model. Hence, we can interpret regression coefficients intuitively and also improve the performance by avoiding overfitting. We analyze theoretical property of IILasso and show that the proposed method is much advantageous for its sign recovery and achieves almost minimax optimal convergence rate. Synthetic and real data analyses also indicate the effectiveness of IILasso.

We study – both in theory and practice– the use of momentum motions in classic iterative hard thresholding (IHT) methods. By simply modifying plain IHT, we investigate its convergence behavior on convex optimization criteria with non-convex constraints, under standard assumptions. In diverse scenaria, we observe that acceleration in IHT leads to significant improvements, compared to state of the art projected gradient descent and Frank-Wolfe variants. As a byproduct of our inspection, we study the impact of selecting the momentum parameter: similar to convex settings, two modes of behavior are observed –“rippling” and linear– depending on the level of momentum.

Many recommendation algorithms rely on user data to generate recommendations. However, these recommendations also affect the data obtained from future users. This work aims to understand the effects of this dynamic interaction. We propose a simple model where users with heterogeneous preferences arrive over time. Based on this model, we prove that naive estimators, i.e. those which ignore this feedback loop, are not consistent. We show that consistent estimators are efficient in the presence of myopic agents. Our results are validated using extensive simulations.

Bayesian optimization (BO) is a popular technique for sequential black-box function optimization, with applications including parameter tuning, robotics, environmental monitoring, and more. One of the most important challenges in BO is the development of algorithms that scale to high dimensions, which remains a key open problem despite recent progress. In this paper, we consider the approach of Kandasamy et al. (2015), in which the high-dimensional function decomposes as a sum of lower-dimensional functions on subsets of the underlying variables. In particular, we significantly generalize this approach by lifting the assumption that the subsets are disjoint, and consider additive models with arbitrary overlap among the subsets. By representing the dependencies via a graph, we deduce an efficient message passing algorithm for optimizing the acquisition function. In addition, we provide an algorithm for learning the graph from samples based on Gibbs sampling. We empirically demonstrate the effectiveness of our methods on both synthetic and real-world data.

In this article, we develop and analyze a homotopy continuation method, referred to as HONES , for solving the sequential generalized projections in Online Newton Step (Hazan et al., 2006b), as well as the generalized problem known as sequential standard quadratic programming. HONES is fast, tuning-free, error-free (up to machine error) and adaptive to the solution sparsity. This is confirmed by both careful theoretical analysis and extensive experiments on both synthetic and real data.

In this paper, we propose a novel sufficient decrease technique for stochastic variance reduced gradient descent methods such as SVRG and SAGA. In order to make sufficient decrease for stochastic optimization, we design a new sufficient decrease criterion, which yields sufficient decrease versions of stochastic variance reduction algorithms such as SVRG-SD and SAGA-SD as a byproduct. We introduce a coefficient to scale current iterate and to satisfy the sufficient decrease property, which takes the decisions to shrink, expand or even move in the opposite direction, and then give two specific update rules of the coefficient for Lasso and ridge regression. Moreover, we analyze the convergence properties of our algorithms for strongly convex problems, which show that our algorithms attain linear convergence rates. We also provide the convergence guarantees of our algorithms for non-strongly convex problems. Our experimental results further verify that our algorithms achieve significantly better performance than their counterparts.

Online portfolio selection research has so far focused mainly on minimizing regret defined in terms of wealth growth. Practical financial decision making, however, is deeply concerned with both wealth and risk. We consider online learning of portfolios of stocks whose prices are governed by arbitrary (unknown) stationary and ergodic processes, where the goal is to maximize wealth while keeping the conditional value at risk (CVaR) below a desired threshold. We characterize the asymptomatically optimal risk-adjusted performance and present an investment strategy whose portfolios are guaranteed to achieve the asymptotic optimal solution while fulfilling the desired risk constraint. We also numerically demonstrate and validate the viability of our method on standard datasets.

The postulate of independence of cause and mechanism (ICM) has recently led to several new causal discovery algorithms. The interpretation of independence and the way it is utilized, however, varies across these methods. Our aim in this paper is to propose a group theoretic framework for ICM to unify and generalize these approaches. In our setting, the cause-mechanism relationship is assessed by perturbing it with random group transformations. We show that the group theoretic view encompasses previous ICM approaches and provides a very general tool to study the structure of data generating mechanisms with direct applications to machine learning.

A common challenge in estimating parameters of probability density functions is the intractability of the normalizing constant. While in such cases maximum likelihood estimation may be implemented using numerical integration, the approach becomes computationally intensive. In contrast, the score matching method of Hyvärinen (2005) avoids direct calculation of the normalizing constant and yields closed-form estimates for exponential families of continuous distributions over $\mathbb{R}^m$. Hyvärinen (2007) extended the approach to distributions supported on the non-negative orthant $\mathbb{R}_+^m$. In this paper, we give a generalized form of score matching for non-negative data that improves estimation efficiency. We also generalize the regularized score matching method of Lin et al. (2016) for non-negative Gaussian graphical models, with improved theoretical guarantees.

We propose a new technique that boosts the convergence of training generative adversarial networks. Generally, the rate of training deep models reduces severely after multiple iterations. A key reason for this phenomenon is that a deep network is expressed using a highly non-convex finite-dimensional model, and thus the parameter gets stuck in a local optimum. Because of this, methods often suffer not only from degeneration of the convergence speed but also from limitations in the representational power of the trained network. To overcome this issue, we propose an additional layer called the gradient layer to seek a descent direction in an infinite-dimensional space. Because the layer is constructed in the infinite-dimensional space, we are not restricted by the specific model structure of finite-dimensional models. As a result, we can get out of the local optima in finite-dimensional models and move towards the global optimal function more directly. In this paper, this phenomenon is explained from the functional gradient method perspective of the gradient layer. Interestingly, the optimization procedure using the gradient layer naturally constructs the deep structure of the network. Moreover, we demonstrate that this procedure can be regarded as a discretization method of the gradient flow that naturally reduces the objective function. Finally, the method is tested using several numerical experiments, which show its fast convergence.

It has been experimentally observed that distributed implementations of mini-batch stochastic gradient descent (SGD) algorithms exhibit speedup saturation and decaying generalization ability beyond a particular batch-size. In this work, we present an analysis hinting that high similarity between concurrently processed gradients may be a cause of this performance degradation. We introduce the notion of gradient diversity that measures the dissimilarity between concurrent gradient updates, and show its key role in the convergence and generalization performance of mini-batch SGD. We also establish that heuristics similar to DropConnect, Langevin dynamics, and quantization, are provably diversity-inducing mechanisms, and provide experimental evidence indicating that these mechanisms can indeed enable the use of larger batches without sacrificing accuracy and lead to faster training in distributed learning. For example, in one of our experiments, for a convolutional neural network to reach 95% training accuracy on MNIST, using the diversity-inducing mechanism can reduce the training time by 30% in the distributed setting.

In high dimension, it is customary to consider Lasso-type estimators to enforce sparsity. For standard Lasso theory to hold, the regularization parameter should be proportional to the noise level, which is often unknown in practice. A remedy is to consider estimators such as the Concomitant Lasso, which jointly optimize over the regression coefficients and the noise level. However, when data from different sources are pooled to increase sample size, noise levels differ and new dedicated estimators are needed. We provide new statistical and computational solutions to perform heteroscedastic regression, with an emphasis on functional brain imaging with magneto- and electroencephalography (M/EEG). When instantiated to de-correlated noise, our framework leads to an efficient algorithm whose computational cost is not higher than for the Lasso, but addresses more complex noise structures. Experiments demonstrate improved prediction and support identification with correct estimation of noise levels.