Toward Opening Book Learning¹

Article type: Research Article

Authors: Buro, M.^{; 2}

Affiliations: NEC Research Institute, Princeton, USA

Note: [1] This article is an improved and updated version of a paper presented at the IJCAI-97 Workshop on Computer Games on Using Games as an Experimental Testbed for AI Research (ed. H. Iida), pp. 1-7. Nagoya, Japan. The current article will be published in a book titled Games in AI Research. The article is presented here as additional information for our readers on book learning after the publication of Hyatt’s article Book Learning – a Methodology to Tune an Opening Book Automatically in the previous issue.

Note: [2] NEC Research Institute, 4 Independence Way, Princeton NJ 08540, USA. Email: mic@research.nj.nec.com.

Abstract: In this contribution an opening-book framework for game-playing programs is presented. The research is motivated by the aspiration to play a sequence of games successfully, i.e., to avoid losing a game twice in the same way. We show how reasonable move alternatives can be found to deviate from previous lines of play. Variants of the algorithm are used by several of today’s best Othello programs to extend their opening books automatically.

DOI: 10.3233/ICG-1999-22204

Journal: ICGA Journal, vol. 22, no. 2, pp. 98-102, 1999