Mean rewards of actions are often correlated. The form of these correlations may be complex and unknown a priori, such as the preferences of a user for recommended products and their categories. To maximize statistical efficiency, it is important to leverage these correlations when learning. We formulate a bandit variant of this problem where the correlations of mean action rewards are represented by a hierarchical Bayesian model with latent variables. Since the hierarchy can have multiple layers, we call it deep. We propose a hierarchical Thompson sampling algorithm (HierTS) for this problem, and show how to implement it efficiently for Gaussian hierarchies. The efficient implementation is possible due to a novel exact hierarchical representation of the posterior, which itself is of independent interest. We use this exact posterior to analyze the Bayes regret of HierTS in Gaussian bandits. Our analysis reflects the structure of the problem, that the regret decreases with the prior width, and also shows that hierarchies reduce the regret by non-constant factors in the number of actions. We confirm these theoretical findings empirically, in both synthetic and real-world experiments.
Meta-, multi-task, and federated learning can be all viewed as solving similar tasks, drawn from an unknown distribution that reflects task similarities. In this work, we provide a unified view of all these problems, as learning to act in a hierarchical Bayesian bandit. We analyze a natural hierarchical Thompson sampling algorithm (hierTS) that can be applied to any problem in this class. Our regret bounds hold under many instances of such problems, including when the tasks are solved sequentially or in parallel; and capture the structure of the problems, such that the regret decreases with the width of the task prior. Our proofs rely on novel total variance decompositions, which can be applied to other graphical model structures. Finally, our theory is complemented by experiments, which show that the hierarchical structure helps with knowledge sharing among the tasks. This confirms that hierarchical Bayesian bandits are a universal and statistically-efficient tool for learning to act with similar bandit tasks.
We study Thompson sampling (TS) in online decision-making problems where the uncertain environment is sampled from a mixture distribution. This is relevant to multi-task settings, where a learning agent is faced with different classes of problems. We incorporate this structure in a natural way by initializing TS with a mixture prior -- dubbed MixTS -- and develop a novel, general technique for analyzing the regret of TS with such priors. We apply this technique to derive Bayes regret bounds for MixTS in both linear bandits and tabular Markov decision processes (MDPs). Our regret bounds reflect the structure of the problem and depend on the number of components and confidence width of each component of the prior. Finally, we demonstrate the empirical effectiveness of MixTS in both synthetic and real-world experiments.
Users of recommender systems often behave in a non-stationary fashion, due to their evolving preferences and tastes over time. In this work, we propose a practical approach for fast personalization to non-stationary users. The key idea is to frame this problem as a latent bandit, where the prototypical models of user behavior are learned offline and the latent state of the user is inferred online from its interactions with the models. We call this problem a non-stationary latent bandit. We propose Thompson sampling algorithms for regret minimization in non-stationary latent bandits, analyze them, and evaluate them on a real-world dataset. The main strength of our approach is that it can be combined with rich offline-learned models, which can be misspecified, and are subsequently fine-tuned online using posterior sampling. In this way, we naturally combine the strengths of offline and online learning.
In many sequence learning tasks, such as program synthesis and document summarization, a key problem is searching over a large space of possible output sequences. We propose to learn representations of the outputs that are specifically meant for search: rich enough to specify the desired output but compact enough to make search more efficient. Discrete latent codes are appealing for this purpose, as they naturally allow sophisticated combinatorial search strategies. The latent codes are learned using a self-supervised learning principle, in which first a discrete autoencoder is trained on the output sequences, and then the resulting latent codes are used as intermediate targets for the end-to-end sequence prediction task. Based on these insights, we introduce the \emph{Latent Programmer}, a program synthesis method that first predicts a discrete latent code from input/output examples, and then generates the program in the target language. We evaluate the Latent Programmer on two domains: synthesis of string transformation programs, and generation of programs from natural language descriptions. We demonstrate that the discrete latent representation significantly improves synthesis accuracy.
A latent bandit problem is one in which the learning agent knows the arm reward distributions conditioned on an unknown discrete latent state. The primary goal of the agent is to identify the latent state, after which it can act optimally. This setting is a natural midpoint between online and offline learning---complex models can be learned offline with the agent identifying latent state online---of practical relevance in, say, recommender systems. In this work, we propose general algorithms for this setting, based on both upper confidence bounds (UCBs) and Thompson sampling. Our methods are contextual and aware of model uncertainty and misspecification. We provide a unified theoretical analysis of our algorithms, which have lower regret than classic bandit policies when the number of latent states is smaller than actions. A comprehensive empirical study showcases the advantages of our approach.
Off-policy learning is a framework for evaluating and optimizing policies without deploying them, from data collected by another policy. Real-world environments are typically non-stationary and the offline learned policies should adapt to these changes. To address this challenge, we study the novel problem of off-policy optimization in piecewise-stationary contextual bandits. Our proposed solution has two phases. In the offline learning phase, we partition logged data into categorical latent states and learn a near-optimal sub-policy for each state. In the online deployment phase, we adaptively switch between the learned sub-policies based on their performance. This approach is practical and analyzable, and we provide guarantees on both the quality of off-policy optimization and the regret during online deployment. To show the effectiveness of our approach, we compare it to state-of-the-art baselines on both synthetic and real-world datasets. Our approach outperforms methods that act only on observed context.
We focus on the problem of predicting future states of entities in complex, real-world driving scenarios. Previous research has used low-level signals to predict short time horizons, and has not addressed how to leverage key assets relied upon heavily by industry self-driving systems: (1) large 3D perception efforts which provide highly accurate 3D states of agents with rich attributes, and (2) detailed and accurate semantic maps of the environment (lanes, traffic lights, crosswalks, etc). We present a unified representation which encodes such high-level semantic information in a spatial grid, allowing the use of deep convolutional models to fuse complex scene context. This enables learning entity-entity and entity-environment interactions with simple, feed-forward computations in each timestep within an overall temporal model of an agent's behavior. We propose different ways of modelling the future as a distribution over future states using standard supervised learning. We introduce a novel dataset providing industry-grade rich perception and semantic inputs, and empirically show we can effectively learn fundamentals of driving behavior.
The evolution of the internet has created an abundance of unstructured data on the web, a significant part of which is textual. The task of author profiling seeks to find the demographics of people solely from their linguistic and content-based features in text. The ability to describe traits of authors clearly has applications in fields such as security and forensics, as well as marketing. Instead of seeing age as just a classification problem, we also frame age as a regression one, but use an ensemble chain method that incorporates the power of both classification and regression to learn the authors exact age.