Playing Games with the Job-Level Computation System

• Item type: Symposium(1)
• 公開日: 2018-11-05
• タイトル: Playing Games with the Job-Level Computation System
• タイトル / 言語: en
• タイトル: Playing Games with the Job-Level Computation System
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: Computer games
• キーワード / 主題Scheme: Other
• キーワード / 主題: Job-level computing
• キーワード / 主題Scheme: Other
• キーワード / 主題: Distributed computing
• キーワード / 主題Scheme: Other
• キーワード / 主題: Connect6
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: National Chiao Tung University
• 著者所属(英): en
• 著者所属(英): National Chiao Tung University
• 著者名: Chung-Chin, Shih; Ting, han Wei; Zheng-Yuan, Lee; I-Chen, Wu
• 著者名(英): Chung-Chin, Shih; Ting, han Wei; Zheng-Yuan, Lee; I-Chen, Wu
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper describes how the job-level computation system is used as a game playing agent, with a 90% win rate for Connect6 tournaments on the website Little Golem between the years 2009 to 2018. In addition, we also construct a win/loss database as an add-on to the existing job-level computation system. This database helps save precious computing resource when encountering previously solved game positions. In the experiments, a benchmark of 32 Connect6 openings is used. The results show that there are about 94.2% and 96.5% of the jobs can be conserved in JL-PNS and JL-UCT respectively for the benchmark when applying the win/loss database. Moreover, we discuss the issues about Graph Interaction Problem and how we generate hash keys of win/loss database.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: This paper describes how the job-level computation system is used as a game playing agent, with a 90% win rate for Connect6 tournaments on the website Little Golem between the years 2009 to 2018. In addition, we also construct a win/loss database as an add-on to the existing job-level computation system. This database helps save precious computing resource when encountering previously solved game positions. In the experiments, a benchmark of 32 Connect6 openings is used. The results show that there are about 94.2% and 96.5% of the jobs can be conserved in JL-PNS and JL-UCT respectively for the benchmark when applying the win/loss database. Moreover, we discuss the issues about Graph Interaction Problem and how we generate hash keys of win/loss database.
• 書誌情報: ゲームプログラミングワークショップ2018論文集; 巻 2018, p. 192-198, 発行日 2018-11-09
• 出版者 / 言語: ja
• 出版者: 情報処理学会