A Comparative Study of Parallel Reinforcement Learning Methods with a PC Cluster System

  • Authors:

  • Masayuki Kushida;Kenichi Takahashi;Hiroaki Ueda;Tetsuhiro Miyahara

  • Affiliations:

  • Hiroshima City University, Japan;Hiroshima City University, Japan;Hiroshima City University, Japan;Hiroshima City University, Japan

  • Venue:
  • IAT '06 Proceedings of the IEEE/WIC/ACM international conference on Intelligent Agent Technology
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a comparative study of three parallel implementation models for reinforcement learning. Two of them utilize Q-learning, and the other one utilizes fuzzy Q-learning for agent learning. In order to evaluate performance and validity of the three method, a PC(personal computer) cluster system consisting of 16 PCs connected via Gigabit ethernet has been built. For communications to deliver data among PCs, MPI (Message Passing Interface) is employed. Experimental results are compared with one another to show the performance and characteristics of the three methods.