Saved in:
Bibliographic Details
Main Authors: Liu, Xinyu, Xie, Zixuan, Zhang, Shangtong
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.19254
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916761010438144
author Liu, Xinyu
Xie, Zixuan
Zhang, Shangtong
author_facet Liu, Xinyu
Xie, Zixuan
Zhang, Shangtong
contents $Q$-learning is one of the most fundamental reinforcement learning algorithms. It is widely believed that $Q$-learning with linear function approximation (i.e., linear $Q$-learning) suffers from possible divergence until the recent work Meyn (2024) which establishes the ultimate almost sure boundedness of the iterates of linear $Q$-learning. Building on this success, this paper further establishes the first $L^2$ convergence rate of linear $Q$-learning iterates (to a bounded set). Similar to Meyn (2024), we do not make any modification to the original linear $Q$-learning algorithm, do not make any Bellman completeness assumption, and do not make any near-optimality assumption on the behavior policy. All we need is an $ε$-softmax behavior policy with an adaptive temperature. The key to our analysis is the general result of stochastic approximations under Markovian noise with fast-changing transition functions. As a side product, we also use this general result to establish the $L^2$ convergence rate of tabular $Q$-learning with an $ε$-softmax behavior policy, for which we rely on a novel pseudo-contraction property of the weighted Bellman optimality operator.
format Preprint
id arxiv_https___arxiv_org_abs_2501_19254
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Linear $Q$-Learning Does Not Diverge in $L^2$: Convergence Rates to a Bounded Set
Liu, Xinyu
Xie, Zixuan
Zhang, Shangtong
Machine Learning
Artificial Intelligence
$Q$-learning is one of the most fundamental reinforcement learning algorithms. It is widely believed that $Q$-learning with linear function approximation (i.e., linear $Q$-learning) suffers from possible divergence until the recent work Meyn (2024) which establishes the ultimate almost sure boundedness of the iterates of linear $Q$-learning. Building on this success, this paper further establishes the first $L^2$ convergence rate of linear $Q$-learning iterates (to a bounded set). Similar to Meyn (2024), we do not make any modification to the original linear $Q$-learning algorithm, do not make any Bellman completeness assumption, and do not make any near-optimality assumption on the behavior policy. All we need is an $ε$-softmax behavior policy with an adaptive temperature. The key to our analysis is the general result of stochastic approximations under Markovian noise with fast-changing transition functions. As a side product, we also use this general result to establish the $L^2$ convergence rate of tabular $Q$-learning with an $ε$-softmax behavior policy, for which we rely on a novel pseudo-contraction property of the weighted Bellman optimality operator.
title Linear $Q$-Learning Does Not Diverge in $L^2$: Convergence Rates to a Bounded Set
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2501.19254