open access

https://hdl.handle.net/2115/14563

An Improvement of Reinforcement Learning Agent’s Policy by using a Bayesian Network

日本語

2003-11

情報処理学会

情報処理学会論文誌

44(11)

2884-2894

Reinforcement learning is a kind of machine learning. It aims to optimize an agent’s policy by adapting the agent to a given environment according to rewards. In this paper, we propose a method for improving policies by using knowledge, in which reinforcement learning agents obtain. We use a Bayesian Network as knowledge of an agent. Its structure is decided by a model selection method based on information theory using series of an agent’s inputoutput and rewards as sample data. A Bayesian Network constructed in our study represents stochastic dependences between input-output and rewards. In our proposed method, policies are improved by supervised learning using the structure of Bayesian Network (i.e. stochastic knowledge). Introducing the mechanism of improving policies makes reinforcement learning agents acquire more effective policies. We carry out simulations in the pursuit problem in order to discuss the characteristics of our proposed method. Furthermore, we discuss the information about agents' policies represented by the Bayesian Network system.

学術雑誌論文

Accepted Manuscript

The copyright of this material is retained by the Information Processing Society of Japan (IPSJ). This material is published on this web site with the agreement of the author (s) and the IPSJ. Please be complied with Copyright Law of Japan and the Code of Ethics of the IPSJ if any users wish to reproduce, make derivative work, distribute or make available to the public any part or whole thereof.

0387-5806