A Recursive Classifier System for Partially Observable Environments

Article type: Research Article

Authors: Hamzeh, Ali | Hashemi, Sattar | Sami, Ashkan | Rahmani, Adel

Affiliations: CSE and IT Department School of Electrical and Computer Engineering Shiraz University, Iran. E-mail: [email protected], [email protected], [email protected] | Computer Engineering Department Iran University of Science and Technology, Iran. E-mail: [email protected]

Abstract: Previously we introduced Parallel Specialized XCS (PSXCS), a distributed-architecture classifier system that detects aliased environmental states and assigns their handling to created subordinate XCS classifier systems. PSXCS uses a history-window approach, but with novel efficiency since the subordinateXCSs, which employ the windows, are only spawned for parts of the state space that are actually aliased. However, because the window lengths are finite and set manually, PSXCS may fail to be optimal in difficult test mazes. This paper introduces Recursive PSXCS (RPSXCS) that automatically spawns windows wherever more history is required. Experimental results show that RPSXCS is both more powerful and learns faster than PSXCS. The present research suggests new potential for history approaches to partially observable environments.

Keywords: Learning Classifier Systems, Partially Observable Environments, XCS, CSXCS

DOI: 10.3233/FI-2009-191

Journal: Fundamenta Informaticae, vol. 97, no. 1-2, pp. 15-40, 2009