Recent work on policy learning from observational data has highlighted the importance of efficient policy evaluation and has proposed reductions to weighted (cost-sensitive) classification. But, efficient policy evaluation need not yield efficient estimation of policy parameters. We consider the estimation problem given by a weighted surrogate-loss classification reduction of policy learning with any score function, either direct, inverse-propensity weighted, or doubly robust. We show that, under a correct specification assumption, the weighted classification formulation need not be efficient for policy parameters. We draw a contrast to actual (possibly weighted) binary classification, where correct specification implies a parametric model, while for policy learning it only implies a semiparametric model. In light of this, we instead propose an estimation approach based on generalized method of moments, which is efficient for the policy parameters. We propose a particular method based on recent developments on solving moment problems using neural networks and demonstrate the efficiency and regret benefits of this method empirically.

In this paper we study the well-known stochastic linear bandit problem where a decision-maker sequentially chooses among a set of given actions in R^d, observes their noisy reward, and aims to maximize her cumulative expected reward over a horizon of length T. We introduce a general family of algorithms for the problem and prove that they are rate optimal. We also show that several well-known algorithms for the problem such as optimism in the face of uncertainty linear bandit (OFUL) and Thompson sampling (TS) are special cases of our family of algorithms. Therefore, we obtain a unified proof of rate optimality for both of these algorithms. Our results include both adversarial action sets (when actions are potentially selected by an adversary) and stochastic action sets (when actions are independently drawn from an unknown distribution). In terms of regret, our results apply to both Bayesian and worst-case regret settings. Our new unified analysis technique also yields a number of new results and solves two open problems known in the literature. Most notably, (1) we show that TS can incur a linear worst-case regret, unless it uses inflated (by a factor of $\sqrt{d}$) posterior variances at each step. This shows that the best known worst-case regret bound for TS, that is given by (Agrawal & Goyal, 2013; Abeille et al., 2017) and is worse (by a factor of \sqrt(d)) than the best known Bayesian regret bound given by Russo and Van Roy (2014) for TS, is tight. This settles an open problem stated in Russo et al., 2018. (2) Our proof also shows that TS can incur a linear Bayesian regret if it does not use the correct prior or noise distribution. (3) Under a generalized gap assumption and a margin condition, as in Goldenshluger & Zeevi, 2013, we obtain a poly-logarithmic (in $T$) regret bound for OFUL and TS in the stochastic setting.

In this paper, we consider the problem of predicting observations generated online by an unknown, partially observed linear system, which is driven by stochastic noise. For such systems the optimal predictor in the mean square sense is the celebrated Kalman filter, which can be explicitly computed when the system model is known. When the system model is unknown, we have to learn how to predict observations online based on finite data, suffering possibly a non-zero regret with respect to the Kalman filter's prediction. We show that it is possible to achieve a regret of the order of $\mathrm{poly}\log(N)$ with high probability, where $N$ is the number of observations collected. Our work is the first to provide logarithmic regret guarantees for the widely used Kalman filter. This is achieved using an online least-squares algorithm, which exploits the approximately linear relation between future observations and past observations. The regret analysis is based on the stability properties of the Kalman filter, recent statistical tools for finite sample analysis of system identification, and classical results for the analysis of least-squares algorithms for time series. Our regret analysis can also be applied for state prediction of the hidden state, in the case of unknown noise statistics but known state-space basis. A fundamental technical contribution is that our bounds hold even for the class of non-explosive systems, which includes the class of marginally stable systems, which was an open problem for the case of online prediction under stochastic noise.

Recently, it has been shown that carefully designed reinforcement learning (RL) algorithms can achieve near-optimal regret in the episodic or the average-reward setting. However, in practice, RL algorithms are applied mostly to the infinite-horizon discounted-reward setting, so it is natural to ask what the lowest regret an algorithm can achieve is in this case, and how close to the optimal the regrets of existing RL algorithms are. In this paper, we prove a regret lower bound of $\Omega\left(\frac{\sqrt{SAT}}{1 - \gamma} - \frac{1}{(1 - \gamma)^2}\right)$ when $T\geq SA$ on any learning algorithm for infinite-horizon discounted Markov decision processes (MDP), where $S$ and $A$ are the numbers of states and actions, $T$ is the number of actions taken, and $\gamma$ is the discounting factor. We also show that a modified version of the double Q-learning algorithm gives a regret upper bound of $\tilde{O}\left(\frac{\sqrt{SAT}}{(1 - \gamma)^{2.5}}\right)$ when $T\geq SA$. Compared to our bounds, previous best lower and upper bounds both have worse dependencies on $T$ and $\gamma$, while our dependencies on $S, A, T$ are optimal. The proof of our upper bound is inspired by recent advances in the analysis of Q-learning in the episodic setting, but the cyclic nature of infinite-horizon MDPs poses many new challenges.

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 step 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 in the update during the test time.

Deep neural networks for classification of videos, 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 the form of optical flow or implicitly by various differences between adjacent frames. In this work we present a manipulation scheme for fooling video classifiers by introducing a spatial patternless temporal perturbation that is practically unnoticed by human observers and undetectable by leading image adversarial pattern detection algorithms. After demonstrating the manipulation of action classification of single videos, we generalize the procedure to make adversarial patterns with temporal invariance that generalizes across different classes for both targeted and untargeted attacks.

Bayesian optimisation is a powerful method for non-convex black-box optimization in low data regimes. However, the question of establishing tight upper bounds for common algorithms in the noiseless setting remains a largely open question. In this paper, we establish new and tightest bounds for two algorithms, namely GP-UCB and Thompson sampling, under the assumption that the objective function is smooth in terms of having a bounded norm in a Mat\'ern RKHS. Importantly, unlike several related works, we do not consider perfect knowledge of the kernel of the Gaussian process emulator used within the Bayesian optimization loop. This allows us to provide results for practical algorithms that sequentially estimate the Gaussian process kernel parameters from the available data.

Deploying the multi-relational tensor structure of a high dimensional feature space, more efficiently improves the performance of machine learning algorithms. One encounters the \emph{curse of dimensionality}, and working with vectorized data fails to preserve the data structure. To mitigate the nonlinear relationship of tensor data more economically, we propose the \emph{Tensor Train Multi-way Multi-level Kernel (TT-MMK)}. This technique combines kernel filtering of the initial input data (\emph{Kernelized Tensor Train (KTT)}), stable reparametrization of the KTT in the Canonical Polyadic (CP) format, and the Dual Structure-preserving Support Vector Machine (\emph{SVM}) Kernel for revealing nonlinear relationships. We demonstrate numerically that the TT-MMK method is more reliable computationally, is less sensitive to tuning parameters, and gives higher prediction accuracy in the SVM classification compared to similar tensorised SVM methods.

Data analysis based on linear methods, which look for correlations between the features describing samples in a data set, or between features and properties associated with the samples, constitute the simplest, most robust, and transparent approaches to the automatic processing of large amounts of data for building supervised or unsupervised machine learning models. Principal covariates regression (PCovR) is an under-appreciated method that interpolates between principal component analysis and linear regression, and can be used to conveniently reveal structure-property relations in terms of simple-to-interpret, low-dimensional maps. Here we provide a pedagogic overview of these data analysis schemes, including the use of the kernel trick to introduce an element of non-linearity in the process, while maintaining most of the convenience and the simplicity of linear approaches. We then introduce a kernelized version of PCovR and a sparsified extension, followed by a feature-selection scheme based on the CUR matrix decomposition modified to incorporate the same hybrid loss that underlies PCovR. We demonstrate the performance of these approaches in revealing and predicting structure-property relations in chemistry and materials science.

In this work, we explore the dependencies between speaker recognition and emotion recognition. We first show that knowledge learned for speaker recognition can be reused for emotion recognition through transfer learning. Then, we show the effect of emotion on speaker recognition. For emotion recognition, we show that using a simple linear model is enough to obtain good performance on the features extracted from pre-trained models such as the x-vector model. Then, we improve emotion recognition performance by fine-tuning for emotion classification. We evaluated our experiments on three different types of datasets: IEMOCAP, MSP-Podcast, and Crema-D. By fine-tuning, we obtained 30.40%, 7.99%, and 8.61% absolute improvement on IEMOCAP, MSP-Podcast, and Crema-D respectively over baseline model with no pre-training. Finally, we present results on the effect of emotion on speaker verification. We observed that speaker verification performance is prone to changes in test speaker emotions. We found that trials with angry utterances performed worst in all three datasets. We hope our analysis will initiate a new line of research in the speaker recognition community.