Winning against Go masters with massively parallel search algorithm using 10,000 or more CPU cores

2016 Fiscal Year Final Research Report

• Project/Area Number: 25700038
• Research Category: Grant-in-Aid for Young Scientists (A)
• Allocation Type: Partial Multi-year Fund
• Research Field: Entertainment and game informatics 1
• Research Institution: Institute of Physical and Chemical Research (2016); Tokyo Institute of Technology (2013-2015)
• Principal Investigator: Yoshizoe Kazuki 国立研究開発法人理化学研究所, 革新知能統合研究センター, ユニットリーダー (80449115)
• Project Period (FY): 2013-04-01 – 2017-03-31
• Keywords: ゲームプログラミング / 並列計算 / 探索 / 人工知能 / 囲碁

Outline of Final Research Achievements

Graph Search problem aims at finding node(s) or path(s) from given graphs, which has various applications including path finding and combinatorial optimization. One of the interesting example is game AI. In the original research proposal, one of our goals was to develop a strong Go program for promoting the research results. However, as is well known, a strong Go program was developed by other research groups and we had focused on other goals, which is improving the performance of parallel search algorithms and applying to various problems. For depth-first search algorithms and hash-table based algorithms, we achieved promising results. Especially for depth-first search applications, we obtained a extremely efficient parallel algorithm for realistic problems.

Free Research Field

探索アルゴリズムと並列計算