Modifications of UCT and sequence-like simulations for Monte-Carlo Go

Author

Wang, Yizao ; Gelly, Sylvain

Author_Institution

Center of Appl. Math., Ecole Polytechnique, Palaiseau

fYear

2007

fDate

1-5 April 2007

Firstpage

175

Lastpage

182

Abstract

Algorithm UCB1 for multi-armed bandit problem has already been extended to algorithm UCT which works for minimax tree search. We have developed a Monte-Carlo program, MoGo, which is the first computer Go program using UCT. We explain our modification of UCT for Go application and also the sequence-like random simulation with patterns which has improved significantly the performance of MoGo. UCT combined with pruning techniques for large Go board is discussed, as well as parallelization of UCT. MoGo is now a top-level computer-Go program on 9 times 9 Go board

Keywords

Monte Carlo methods; computer games; games of skill; Go board; MoGo; Monte-Carlo Go; Monte-Carlo program; algorithm UCB1; algorithm UCT; computer Go program; level computer-Go program; minimax tree search; multiarmed bandit problem; pruning technique; sequence-like random simulation; Application software; Computational intelligence; Computational modeling; Gold; Heuristic algorithms; Mathematics; Medals; Minimax techniques; Stochastic processes; Tree data structures; 3 Ã\x97 3 patterns; Computer Go; Monte-Carlo Go; UCT; multi-armed bandit; sequence-like simulation;

fLanguage

English

Publisher

ieee

Conference_Titel

Computational Intelligence and Games, 2007. CIG 2007. IEEE Symposium on

Conference_Location

Honolulu, HI

Print_ISBN

1-4244-0709-5

Type

conf

DOI

10.1109/CIG.2007.368095

Filename

4219040