Projection by Convolution: Optimal Sample Complexity for Reinforcement Learning in Continuous-Space MDPs
Abstract
We consider the problem of learning an -optimal policy in a general class of continuous-space Markov decision processes (MDPs) having smooth Bellman operators. Given access to a generative model, we achieve rate-optimal sample complexity by performing a simple, \emph{perturbed} version of least-squares value iteration with orthogonal trigonometric polynomials as features. Key to our solution is a novel projection technique based on ideas from harmonic analysis. Our~ sample complexity, where is the dimension of the state-action space and the order of smoothness, recovers the state-of-the-art result of discretization approaches for the special case of Lipschitz MDPs . At the same time, for , it recovers and greatly generalizes the rate of low-rank MDPs, which are more amenable to regression approaches. In this sense, our result bridges the gap between two popular but conflicting perspectives on continuous-space MDPs.
Cite
@article{arxiv.2405.06363,
title = {Projection by Convolution: Optimal Sample Complexity for Reinforcement Learning in Continuous-Space MDPs},
author = {Davide Maran and Alberto Maria Metelli and Matteo Papini and Marcello Restelli},
journal= {arXiv preprint arXiv:2405.06363},
year = {2024}
}