Projection-based Lyapunov method for fully heterogeneous weakly-coupled MDPs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Xiangcheng, Hong, Yige, Wang, Weina
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908765853319168
author Zhang, Xiangcheng
Hong, Yige
Wang, Weina
author_facet Zhang, Xiangcheng
Hong, Yige
Wang, Weina
contents Heterogeneity poses a fundamental challenge for many real-world large-scale decision-making problems but remains largely understudied. In this paper, we study the fully heterogeneous setting of a prominent class of such problems, known as weakly-coupled Markov decision processes (WCMDPs). Each WCMDP consists of $N$ arms (or subproblems), which have distinct model parameters in the fully heterogeneous setting, leading to the curse of dimensionality when $N$ is large. We show that, under mild assumptions, an efficiently computable policy achieves an $O(1/\sqrt{N})$ optimality gap in the long-run average reward per arm for fully heterogeneous WCMDPs as $N$ becomes large. This is the first asymptotic optimality result for fully heterogeneous average-reward WCMDPs. Our main technical innovation is the construction of projection-based Lyapunov functions that certify the convergence of rewards and costs to an optimal region, even under full heterogeneity.
format Preprint
id arxiv_https___arxiv_org_abs_2502_06072
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Projection-based Lyapunov method for fully heterogeneous weakly-coupled MDPs
Zhang, Xiangcheng
Hong, Yige
Wang, Weina
Machine Learning
Optimization and Control
Probability
90C40
G.3; I.6
Heterogeneity poses a fundamental challenge for many real-world large-scale decision-making problems but remains largely understudied. In this paper, we study the fully heterogeneous setting of a prominent class of such problems, known as weakly-coupled Markov decision processes (WCMDPs). Each WCMDP consists of $N$ arms (or subproblems), which have distinct model parameters in the fully heterogeneous setting, leading to the curse of dimensionality when $N$ is large. We show that, under mild assumptions, an efficiently computable policy achieves an $O(1/\sqrt{N})$ optimality gap in the long-run average reward per arm for fully heterogeneous WCMDPs as $N$ becomes large. This is the first asymptotic optimality result for fully heterogeneous average-reward WCMDPs. Our main technical innovation is the construction of projection-based Lyapunov functions that certify the convergence of rewards and costs to an optimal region, even under full heterogeneity.
title Projection-based Lyapunov method for fully heterogeneous weakly-coupled MDPs
topic Machine Learning
Optimization and Control
Probability
90C40
G.3; I.6
url https://arxiv.org/abs/2502.06072