Associating domain-dependent knowledge and Monte Carlo approaches within a Go program

Abstract

This paper underlines the association of two computer go approaches, a domain-dependent knowledge approach and Monte Carlo. First, the strengthes and weaknesses of the two existing approaches are related. Then, the association is described in two steps. A first step consists in using domain-dependent knowledge within the random games enabling the program to compute evaluations that are more significant than before. A second step simply lies in pre-processing the Monte Carlo process with a knowledge-based move generator in order to speed up the program and to eliminate tactically bad moves. We set up experiments demonstrating the relevance of this association, used by Indigo at the 8th computer olympiad as well.