Value-Iteration Based Fitted Policy Iteration: Learning with a Single Trajectory

Andras Antos 1 Csaba Szepesvari 2 Rémi Munos 3
1 MTA SZTAKI - Computer and Automation Research Institute [Budapest]
2 Department of Computing Science [Edmonton]
3 SEQUEL - Sequential Learning
LIFL - Laboratoire d'Informatique Fondamentale de Lille, Inria Lille - Nord Europe, LAGIS - Laboratoire d'Automatique, Génie Informatique et Signal
Abstract : We consider batch reinforcement learning problems in continuous space,expected total discounted-reward Markovian Decision Problems when the training data is composed of the trajectory of some fixed behaviour policy. The algorithm studied is policy iteration where in successive iterations the action-value functions of the intermediate policies are obtained by means of approximate value iteration. PAC-style polynomial bounds are derived on the number of samples needed to guarantee near-optimal performance. The bounds depend on the mixing rate of the trajectory, the smoothness properties of the underlying Markovian Decision Problem, the approximation power and capacity of the function set used. One of the main novelties of the paper is that new smoothness constraints are introduced thereby significantly extending the scope of previous results.
Keywords : Reinforcement Learning Markov decision process policy iteration statistical learning function approximation
Document type :
Conference papers
Complete list of metadatas

Cited literature [16 references]  Display  Hide  Download
https://hal.inria.fr/inria-00124833
Contributor : Rémi Munos <>
Submitted on : Tuesday, January 16, 2007 - 1:34:47 PM
Last modification on : Wednesday, December 11, 2019 - 5:10:02 PM
Long-term archiving on: Friday, September 21, 2012 - 10:15:57 AM

Files

sapi_adprl_final.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00124833, version 1

Collections

INRIA | CNRS | UNIV-LILLE3 | LAGIS

Citation

Andras Antos, Csaba Szepesvari, Rémi Munos. Value-Iteration Based Fitted Policy Iteration: Learning with a Single Trajectory. IEEE International Symposium on Approximate Dynamic Programming and Reinforcement Learning, 2007, Hawai, United States. pp.2007. ⟨inria-00124833⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

392

Files downloads

425

 

 

Contact

archive-ouverte@inria.fr