Finite-time Analysis of the Multiarmed Bandit Problem
Machine Learning
CG '08 Proceedings of the 6th international conference on Computers and Games
Monte-Carlo Tree Search Solver
CG '08 Proceedings of the 6th international conference on Computers and Games
Achieving master level play in 9×9 computer go
AAAI'08 Proceedings of the 23rd national conference on Artificial intelligence - Volume 3
Monte Carlo tree search techniques in the game of Kriegspiel
IJCAI'09 Proceedings of the 21st international jont conference on Artifical intelligence
Efficient selectivity and backup operators in Monte-Carlo tree search
CG'06 Proceedings of the 5th international conference on Computers and games
Bandit based monte-carlo planning
ECML'06 Proceedings of the 17th European conference on Machine Learning
Monte-Carlo tree search in settlers of catan
ACG'09 Proceedings of the 12th international conference on Advances in Computer Games
Move-Pruning techniques for monte-carlo go
ACG'05 Proceedings of the 11th international conference on Advances in Computer Games
| Hi-index | 0.00 |
Recent works on the MCTS and UCT framework in the domain of Go focused on introducing knowledge to the playout and on pruning variations from the tree, but so far node expansion has not been investigated. In this paper we show that delaying expansion according to the number of the siblings delivers a gain of more than 92% when compared to normal expansion. We propose three improvements; one that uses domain knowledge and two that are domain-independent methods. Experimental results show that all advanced operators significantly improve the UCT performance when compared to the basic delaying expansion. From the results we may conclude that the new expansion operators are an appropriate means to improve the UCT algorithm.