New Versions of Gradient Temporal Difference Learning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lee, Donghwan, Lim, Han-Dong, Park, Jihoon, Choi, Okyong
Format: Preprint
Veröffentlicht: 2021
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917570729213952
author Lee, Donghwan
Lim, Han-Dong
Park, Jihoon
Choi, Okyong
author_facet Lee, Donghwan
Lim, Han-Dong
Park, Jihoon
Choi, Okyong
contents Sutton, Szepesvári and Maei introduced the first gradient temporal-difference (GTD) learning algorithms compatible with both linear function approximation and off-policy training. The goal of this paper is (a) to propose some variants of GTDs with extensive comparative analysis and (b) to establish new theoretical analysis frameworks for the GTDs. These variants are based on convex-concave saddle-point interpretations of GTDs, which effectively unify all the GTDs into a single framework, and provide simple stability analysis based on recent results on primal-dual gradient dynamics. Finally, numerical comparative analysis is given to evaluate these approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2109_04033
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle New Versions of Gradient Temporal Difference Learning
Lee, Donghwan
Lim, Han-Dong
Park, Jihoon
Choi, Okyong
Machine Learning
Sutton, Szepesvári and Maei introduced the first gradient temporal-difference (GTD) learning algorithms compatible with both linear function approximation and off-policy training. The goal of this paper is (a) to propose some variants of GTDs with extensive comparative analysis and (b) to establish new theoretical analysis frameworks for the GTDs. These variants are based on convex-concave saddle-point interpretations of GTDs, which effectively unify all the GTDs into a single framework, and provide simple stability analysis based on recent results on primal-dual gradient dynamics. Finally, numerical comparative analysis is given to evaluate these approaches.
title New Versions of Gradient Temporal Difference Learning
topic Machine Learning
url https://arxiv.org/abs/2109.04033