Metadata-based Multi-Task Bandits with Bayesian Hierarchical Models
This work addresses the challenge of knowledge transfer in multi-task bandit problems for researchers and practitioners in reinforcement learning, though it appears incremental as it builds on existing Bayesian hierarchical models and Thompson sampling methods.
The paper tackles the problem of efficient exploration in multi-armed bandits by introducing a metadata-based multi-task framework that leverages task-specific features to share knowledge across tasks, resulting in a Thompson sampling algorithm that minimizes cumulative regrets and demonstrates benefits in Gaussian bandits with clear Bayes regret improvements.
How to explore efficiently is a central problem in multi-armed bandits. In this paper, we introduce the metadata-based multi-task bandit problem, where the agent needs to solve a large number of related multi-armed bandit tasks and can leverage some task-specific features (i.e., metadata) to share knowledge across tasks. As a general framework, we propose to capture task relations through the lens of Bayesian hierarchical models, upon which a Thompson sampling algorithm is designed to efficiently learn task relations, share information, and minimize the cumulative regrets. Two concrete examples for Gaussian bandits and Bernoulli bandits are carefully analyzed. The Bayes regret for Gaussian bandits clearly demonstrates the benefits of information sharing with our algorithm. The proposed method is further supported by extensive experiments.