More Efficient Randomized Exploration for Reinforcement Learning via Approximate Sampling

TL;DR

This paper introduces a flexible framework combining approximate sampling methods with Feel-Good Thompson Sampling to improve exploration in reinforcement learning, achieving better regret bounds and empirical performance in deep RL tasks.

Contribution

It develops a novel algorithmic framework that integrates various approximate sampling techniques with FGTS, enhancing exploration efficiency and theoretical guarantees in RL.

Findings

Achieves the best known regret dependency on dimensionality for linear MDPs.

Provides explicit sampling complexity for each sampler used.

Demonstrates superior empirical performance on Atari games.

Abstract

Thompson sampling (TS) is one of the most popular exploration techniques in reinforcement learning (RL). However, most TS algorithms with theoretical guarantees are difficult to implement and not generalizable to Deep RL. While the emerging approximate sampling-based exploration schemes are promising, most existing algorithms are specific to linear Markov Decision Processes (MDP) with suboptimal regret bounds, or only use the most basic samplers such as Langevin Monte Carlo. In this work, we propose an algorithmic framework that incorporates different approximate sampling methods with the recently proposed Feel-Good Thompson Sampling (FGTS) approach (Zhang, 2022; Dann et al., 2021), which was previously known to be computationally intractable in general. When applied to linear MDPs, our regret analysis yields the best known dependency of regret on dimensionality, surpassing existing randomized algorithms. Additionally, we provide explicit sampling complexity for each employed sampler. Empirically, we show that in tasks where deep exploration is necessary, our proposed algorithms that combine FGTS and approximate sampling perform significantly better compared to other strong baselines. On several challenging games from the Atari 57 suite, our algorithms achieve performance that is either better than or on par with other strong baselines from the deep RL literature.