Scalable distributed Monte-Carlo tree search

Kazuki Yoshizoe, Akihiro Kishimoto, Tomoyuki Kaneko, Haruhiro Yoshimoto, Yutaka Ishikawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

27 Citations (Scopus)

Abstract

Monte-Carlo Tree Search (MCTS) is remarkably successful in two-player games, but parallelizing MCTS has been notoriously difficult to scale well, especially in distributed environments. For a distributed parallel search, transposition-table driven scheduling (TDS) is known to be efficient in several domains. We present a massively parallel MCTS algorithm, that applies the TDS parallelism to the Upper Confidence bound Applied to Trees (UCT) algorithm, which is the most representative MCTS algorithm. To drastically decrease communication overhead, we introduce a reformulation of UCT called Depth-First UCT. The parallel performance of the algorithm is evaluated on clusters using up to 1,200 cores in artificial game-trees. We show that this approach scales well, achieving 740-fold speedups in the best case.

Original languageEnglish
Title of host publicationProceedings of the 4th Annual Symposium on Combinatorial Search, SoCS 2011
Pages180-187
Number of pages8
Publication statusPublished - 2011
Externally publishedYes
Event4th International Symposium on Combinatorial Search, SoCS 2011 - Barcelona, Spain
Duration: Jul 15 2011Jul 16 2011

Publication series

NameProceedings of the 4th Annual Symposium on Combinatorial Search, SoCS 2011

Conference

Conference4th International Symposium on Combinatorial Search, SoCS 2011
Country/TerritorySpain
CityBarcelona
Period7/15/117/16/11

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications

Fingerprint

Dive into the research topics of 'Scalable distributed Monte-Carlo tree search'. Together they form a unique fingerprint.

Cite this