EFFICIENT CONTROL OF SELECTIVE SIMULATIONS

Article type: Research Article

Authors: Sheppard, Brian¹

Affiliations: Concord, MA, USA

Note: [1] Sheppard Company, Inc, Concord, MA, USA. E-mail: sheppardco@aol.com

Note: [2] The article is a slightly revised version of the author’s invited address at the CG ’04 Conference in Ramat-Gan, Israel (Sheppard, 2004).

Abstract: Selective simulation is a search technique that estimates the value of a move in a state space by averaging the results of a selected sample of continuations. The value of selective sampling has been demonstrated in domains such as Backgammon, Scrabble, poker, bridge, and even Go. This article describes efficient methods for controlling selective simulations.2

DOI: 10.3233/ICG-2004-27202

Journal: ICGA Journal, vol. 27, no. 2, pp. 67-80, 2004