Finite-Sample Analysis of Least-Squares Policy Iteration - Archive ouverte HAL Accéder directement au contenu
Rapport (Rapport Technique) Année : 2010

Finite-Sample Analysis of Least-Squares Policy Iteration

Alessandro Lazaric
Mohammad Ghavamzadeh
  • Fonction : Auteur
  • PersonId : 868946
Rémi Munos
  • Fonction : Auteur
  • PersonId : 836863

Résumé

In this paper, we report a performance bound for the widely used least-squares policy iteration (LSPI) algorithm. We first consider the problem of policy evaluation in reinforcement learning, i.e., learning the value function of a fixed policy, using the least-squares temporal-difference (LSTD) learning method, and report finite-sample analysis for this algorithm. To do so, we first derive a bound on the performance of the LSTD solution evaluated at the states generated by the Markov chain and used by the algorithm to learn an estimate of the value function. This result is general in the sense that no assumption is made on the existence of a stationary distribution for the Markov chain. We then derive generalization bounds in the case when the Markov chain possesses a stationary distribution and is $\beta$-mixing. Finally, we analyze how the error at each policy evaluation step is propagated through the iterations of a policy iteration method, and derive a performance bound for the LSPI algorithm.

Domaines

Informatique
Fichier principal
Vignette du fichier
lspi-jmlr.pdf (317.34 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

inria-00528596 , version 1 (22-10-2010)

Identifiants

  • HAL Id : inria-00528596 , version 1

Citer

Alessandro Lazaric, Mohammad Ghavamzadeh, Rémi Munos. Finite-Sample Analysis of Least-Squares Policy Iteration. [Technical Report] 2010. ⟨inria-00528596⟩
259 Consultations
108 Téléchargements

Partager

Gmail Facebook X LinkedIn More