Monte-Carlo Proof-Number Search for Computer Go

In the last decade, proof-number search and Monte-Carlo methods have successfully been applied to the combinatorial-games domain. Proof-number search is a reliable algorithm. It requires a well defined goal to prove. This can be seen as a disadvantage. In contrast to proof-number search, Monte-Carlo evaluation is a flexible stochastic evaluation for game-tree search. In order to improve the efficiency of proof-number search, we introduce a new algorithm, Monte-Carlo ProofNumber search. It enhances proof-number search by adding the flexible Monte-Carlo evaluation. We present the new algorithm and evaluate it on a sub-problem of Go, the Life-and-Death problem. The results show a clear improvement in time efficiency and memory usage: the test problems are solved two times faster and four times less nodes are expanded on average. Future work will assess the possibility of applying this method to enhanced proof-number techniques.

• Type: Conference
• Year: 2006
• Where: CG
• Authors: Jahn-Takeshi Saito, Guillaume Chaslot, Jos W. H. M. Uiterwijk, H. Jaap van den Herik
• Keywords: CG 2006, Computer Graphics, Monte-Carlo, Monte-Carlo Evaluation, Proof-number Search