Sample-Efficient Constrained Reinforcement Learning with General Parameterization

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Mondal, Washim Uddin, Aggarwal, Vaneet
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913568234930176
author Mondal, Washim Uddin
Aggarwal, Vaneet
author_facet Mondal, Washim Uddin
Aggarwal, Vaneet
contents We consider a constrained Markov Decision Problem (CMDP) where the goal of an agent is to maximize the expected discounted sum of rewards over an infinite horizon while ensuring that the expected discounted sum of costs exceeds a certain threshold. Building on the idea of momentum-based acceleration, we develop the Primal-Dual Accelerated Natural Policy Gradient (PD-ANPG) algorithm that ensures an $ε$ global optimality gap and $ε$ constraint violation with $\tilde{\mathcal{O}}((1-γ)^{-7}ε^{-2})$ sample complexity for general parameterized policies where $γ$ denotes the discount factor. This improves the state-of-the-art sample complexity in general parameterized CMDPs by a factor of $\mathcal{O}((1-γ)^{-1}ε^{-2})$ and achieves the theoretical lower bound in $ε^{-1}$.
format Preprint
id arxiv_https___arxiv_org_abs_2405_10624
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Sample-Efficient Constrained Reinforcement Learning with General Parameterization
Mondal, Washim Uddin
Aggarwal, Vaneet
Machine Learning
Artificial Intelligence
We consider a constrained Markov Decision Problem (CMDP) where the goal of an agent is to maximize the expected discounted sum of rewards over an infinite horizon while ensuring that the expected discounted sum of costs exceeds a certain threshold. Building on the idea of momentum-based acceleration, we develop the Primal-Dual Accelerated Natural Policy Gradient (PD-ANPG) algorithm that ensures an $ε$ global optimality gap and $ε$ constraint violation with $\tilde{\mathcal{O}}((1-γ)^{-7}ε^{-2})$ sample complexity for general parameterized policies where $γ$ denotes the discount factor. This improves the state-of-the-art sample complexity in general parameterized CMDPs by a factor of $\mathcal{O}((1-γ)^{-1}ε^{-2})$ and achieves the theoretical lower bound in $ε^{-1}$.
title Sample-Efficient Constrained Reinforcement Learning with General Parameterization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2405.10624