Asymptotically optimal regret in communicating Markov decision processes

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Boone, Victor
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910964621770752
author Boone, Victor
author_facet Boone, Victor
contents In this paper, we present a learning algorithm that achieves asymptotically optimal regret for Markov decision processes in average reward under a communicating assumption. That is, given a communicating Markov decision process $M$, our algorithm has regret $K(M) \log(T) + \mathrm{o}(\log(T))$ where $T$ is the number of learning steps and $K(M)$ is the best possible constant. This algorithm works by explicitly tracking the constant $K(M)$ to learn optimally, then balances the trade-off between exploration (playing sub-optimally to gain information), co-exploration (playing optimally to gain information) and exploitation (playing optimally to score maximally). We further show that the function $K(M)$ is discontinuous, which is a consequence challenge for our approach. To that end, we describe a regularization mechanism to estimate $K(M)$ with arbitrary precision from empirical data.
format Preprint
id arxiv_https___arxiv_org_abs_2505_18064
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Asymptotically optimal regret in communicating Markov decision processes
Boone, Victor
Machine Learning
In this paper, we present a learning algorithm that achieves asymptotically optimal regret for Markov decision processes in average reward under a communicating assumption. That is, given a communicating Markov decision process $M$, our algorithm has regret $K(M) \log(T) + \mathrm{o}(\log(T))$ where $T$ is the number of learning steps and $K(M)$ is the best possible constant. This algorithm works by explicitly tracking the constant $K(M)$ to learn optimally, then balances the trade-off between exploration (playing sub-optimally to gain information), co-exploration (playing optimally to gain information) and exploitation (playing optimally to score maximally). We further show that the function $K(M)$ is discontinuous, which is a consequence challenge for our approach. To that end, we describe a regularization mechanism to estimate $K(M)$ with arbitrary precision from empirical data.
title Asymptotically optimal regret in communicating Markov decision processes
topic Machine Learning
url https://arxiv.org/abs/2505.18064