Evaluating Root Parallelization in Go

被引:16
作者
Soejima, Yusuke [1 ]
Kishimoto, Akihiro [1 ,2 ]
Watanabe, Osamu [1 ]
机构
[1] Tokyo Inst Technol, Dept Math & Comp Sci, Tokyo 1528552, Japan
[2] Japan Sci & Technol Agcy, Kawaguchi, Saitama 3320012, Japan
基金
日本科学技术振兴机构; 日本学术振兴会;
关键词
Computer Go; majority voting; Monte Carlo tree search (MCTS); root parallelization; tree parallelization; GAME;
D O I
10.1109/TCIAIG.2010.2096427
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Parallelizing Monte Carlo tree search (MCTS) has been considered to be a way to improve the strength of Computer Go programs. In this paper, we analyze the performance of two root parallelization methods: the standard strategy based on average selection and our new strategy based on majority voting. As a starting code base, we used Fuego, which is one of the best programs available. Our experimental results with 64 central processing unit (CPU) cores show that majority voting outperforms average selection. Additionally, we show through an extensive analysis that root parallelization has limitations.
引用
收藏
页码:278 / 287
页数:10
相关论文
共 23 条
[1]  
[Anonymous], 2006, 6062 INRIA
[2]  
BROCKINGTON M, 1997, ADV COMPUTER CHESS, V8, P69
[3]   A taxonomy of parallel game-tree search algorithms [J].
Brockington, MG .
ICCA JOURNAL, 1996, 19 (03) :162-174
[4]  
Brockington MG, 1998, THESIS U ALBERTA EDM
[5]  
Cazenave T., 2007, P COMP GAM WORKSH, P93
[6]  
Cazenave T, 2008, LECT NOTES COMPUT SC, V5131, P72, DOI 10.1007/978-3-540-87608-3_7
[7]   Parallel Monte-Carlo Tree Search [J].
Chaslot, Guillaume M. J. -B. ;
Winands, Mark H. M. ;
van den Herik, H. Jaap .
COMPUTERS AND GAMES, 2008, 5131 :60-+
[8]   PROGRESSIVE STRATEGIES FOR MONTE-CARLO TREE SEARCH [J].
Chaslot, Guillaume M. J-B. ;
Winands, Mark H. M. ;
Van den Herik, H. Jaap ;
Uiterwijk, Jos W. H. M. ;
Bouzy, Bruno .
NEW MATHEMATICS AND NATURAL COMPUTATION, 2008, 4 (03) :343-357
[9]  
Coulom R, 2007, ICGA J, V30, P198
[10]  
ENZENBERGER M, 2009, 0908 TR U ALB