A Self-Learning, Pattern-Oriented Chess Program

Article type: Research Article

Authors: Levinson, Robert Arlen

Affiliations: University of California, Santa Cruz, USA

Abstract: This contribution describes a research effort to design a pattern-oriented chess program that is to learn to be a good chess-player simply by playing itself. The major results include: (1) the design of an associative retrieval system and representation scheme for chess patterns and positions which can serve as the basis for a precedence finding mechanism similar to that employed by chess masters; (2) the foundations of a learning model based on recent results in machine-learning research which indicate that an adaptive pattern-oriented chess program is possible; (3) a demonstration that search and planning knowledge can be efficiently represented and applied when in the form of patterns.

DOI: 10.3233/ICG-1989-12403

Journal: ICGA Journal, vol. 12, no. 4, pp. 207-215, 1989