Reducing sample complexity in reinforcement learning by transferring transition and reward probabilities

Kouta Oguni, Kazuyuki Narisawa, Ayumi Shinohara

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Most existing reinforcement learning algorithms require many trials until they obtain optimal policies. In this study, we apply transfer learning to reinforcement learning to realize greater efficiency. We propose a new algorithm called TR-MAX, based on the R-MAX algorithm. TR-MAX transfers the transition and reward probabilities from a source task to a target task as prior knowledge. We theoretically analyze the sample complexity of TR-MAX. Moreover, we show that TR-MAX performs much better in practice than R-MAX in maze tasks.

Original languageEnglish
Title of host publicationICAART 2014 - Proceedings of the 6th International Conference on Agents and Artificial Intelligence
PublisherSciTePress
Pages632-638
Number of pages7
ISBN (Print)9789897580154
DOIs
Publication statusPublished - 2014
Event6th International Conference on Agents and Artificial Intelligence, ICAART 2014 - Angers, Loire Valley, France
Duration: 2014 Mar 62014 Mar 8

Publication series

NameICAART 2014 - Proceedings of the 6th International Conference on Agents and Artificial Intelligence
Volume1

Other

Other6th International Conference on Agents and Artificial Intelligence, ICAART 2014
Country/TerritoryFrance
CityAngers, Loire Valley
Period14/3/614/3/8

Keywords

  • PAC-MDP
  • Reinforcement learning
  • Sample complexity
  • Transfer learning

ASJC Scopus subject areas

  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Reducing sample complexity in reinforcement learning by transferring transition and reward probabilities'. Together they form a unique fingerprint.

Cite this