论文信息 - Amazons Discover Monte-Carlo

Amazons Discover Monte-Carlo

Monte-Carlo algorithms and their UCT-like successors have recently shown remarkable promise for Go-playing programs. We apply some of these same algorithms to an Amazons-playing program. Our experiments suggest that a pure MC/UCT type program for playing Amazons has little promise, but by using strong evaluation functions we are able to create a hybrid MC/UCT program that is superior to both the basic MC/UCT program and the conventional minimax-based programs. The MC/UCT program is able to beat Invader , a strong minimax program, over 80% of the time at tournament time controls.

Richard J. Lorentz

[1] Olivier Teytaud,et al. Modification of UCT with Patterns in Monte-Carlo Go , 2006 .

[2] Tristan Cazenave. Iterative Widening , 2001, IJCAI.

[3] Jens Lieberum,et al. An evaluation function for the game of amazons , 2005, Theor. Comput. Sci..

[4] H. Jaap van den Herik,et al. Progressive Strategies for Monte-Carlo Tree Search , 2008 .

[5] Richard J. Lorentz. FIRST-TIME ENTRY AMAZONG WINS AMAZONS TOURNAMENT , 2002 .

[6] Sylvain Gelly,et al. Exploration exploitation in Go: UCT for Monte-Carlo Go , 2006, NIPS 2006.

[7] Henry Avetisyan,et al. Selective Search in an Amazons Program , 2002, Computers and Games.

[8] Zongmin Ma,et al. Computers and Games , 2008, Lecture Notes in Computer Science.

[9] Csaba Szepesvári,et al. Bandit Based Monte-Carlo Planning , 2006, ECML.

[10] Rémi Coulom,et al. Computing "Elo Ratings" of Move Patterns in the Game of Go , 2007, J. Int. Comput. Games Assoc..

[11] Rémi Coulom,et al. Efficient Selectivity and Backup Operators in Monte-Carlo Tree Search , 2006, Computers and Games.