Reinforcement Learning with Logarithmic Regret and Policy SwitchesDownload PDF

Published: 31 Oct 2022, 18:00, Last Modified: 14 Jan 2023, 13:48NeurIPS 2022 AcceptReaders: Everyone
Keywords: reinforcement learning theory, function approximation, instance-dependent regret
TL;DR: We provide instance-dependent regret guarantees for model-based and model-free algorithms in the general function approximation setting, where the underlying function class has bounded eluder dimension.
Abstract: In this paper, we study the problem of regret minimization for episodic Reinforcement Learning (RL) both in the model-free and the model-based setting. We focus on learning with general function classes and general model classes, and we derive results that scale with the eluder dimension of these classes. In contrast to the existing body of work that mainly establishes instance-independent regret guarantees, we focus on the instance-dependent setting and show that the regret scales logarithmically with the horizon $T$, provided that there is a gap between the best and the second best action in every state. In addition, we show that such a logarithmic regret bound is realizable by algorithms with $O(\log T)$ switching cost (also known as adaptivity complexity). In other words, these algorithms rarely switch their policy during the course of their execution. Finally, we complement our results with lower bounds which show that even in the tabular setting, we cannot hope for regret guarantees lower than $O(\log T)$.
Supplementary Material: pdf
11 Replies