{"created":"2026-02-17T06:52:07.104660+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:02007486","sets":["1164:5305:1771205017738:1771205084619"]},"path":["1771205084619"],"owner":"80578","recid":"2007486","title":["強化学習を用いた対戦相手特化ガイスターAIの研究"],"pubdate":{"attribute_name":"PubDate","attribute_value":"2026-02-23"},"_buckets":{"deposit":"9891769e-021c-4abe-a9db-e7718ac38ce6"},"_deposit":{"id":"2007486","pid":{"type":"depid","value":"2007486","revision_id":0},"owners":[80578],"status":"published","created_by":80578},"item_title":"強化学習を用いた対戦相手特化ガイスターAIの研究","author_link":[],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"強化学習を用いた対戦相手特化ガイスターAIの研究","subitem_title_language":"ja"},{"subitem_title":"Research on Geister AI Specialized for Specific Opponents Using Reinforcement Learning","subitem_title_language":"en"}]},"item_type_id":"4","publish_date":"2026-02-23","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"松江工業高等専門学校"},{"subitem_text_value":"松江工業高等専門学校"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"National Institute of Technology, Matsue College","subitem_text_language":"en"},{"subitem_text_value":"National Institute of Technology, Matsue College","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/2007486/files/IPSJ-GI26057019.pdf","label":"IPSJ-GI26057019.pdf"},"date":[{"dateType":"Available","dateValue":"2028-02-23"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-GI26057019.pdf","filesize":[{"value":"652.1 KB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"18"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"e2632cad-b3fa-45eb-8dda-076c54471fa8","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2026 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"長澤,瑛信"}]},{"creatorNames":[{"creatorName":"橋本,剛"}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Eishin Nagasawa","creatorNameLang":"en"}]},{"creatorNames":[{"creatorName":"Tsuyoshi Hashimoto","creatorNameLang":"en"}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA11362144","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-8736","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"完全情報ゲームでは盤面情報がすべて開示されているため，膨大な局面を探索可能なAIが人間を凌駕する性能を示している．一方，不完全情報ゲームでは情報の一部が観測できず，運や推測が勝敗に影響するため，AI開発は困難であり，現在はナッシュ均衡を目指す手法が主流である．しかし，不完全情報ゲームでは対戦相手の癖や戦略に応じて最善手が変化すると考えられる．本研究では，不完全情報ゲームであるガイスターを題材とし，強化学習を用いて特定の対戦相手の癖を捉えるAIを実装した．通常の自己対戦による学習ではなく，特定のAIとの対戦によって得られた棋譜を用いて学習を行うことで，相手の行動傾向を反映した戦略の獲得を目指した．実験の結果，自己対戦によって学習したAIと比較して，特化学習を行ったAIの勝率が向上することを確認した．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"In perfect information games, where all board information is disclosed, AI capable of exploring vast numbers of positions demonstrates performance surpassing humans. Conversely, in imperfect information games, where some information remains unobservable and luck or guesswork influences outcomes, AI development is challenging. Currently, methods aiming for Nash equilibria are mainstream. However, in imperfect information games, the optimal move is thought to change depending on the opponent's habits and strategy. This research uses the imperfect information game Geister as a subject, implementing an AI that captures specific opponent tendencies using reinforcement learning. Rather than learning through standard self-play, it aims to acquire strategies reflecting the opponent's behavioral patterns by learning from game records obtained through matches against a specific AI. Experimental results confirmed that the win rate of the AI trained through specialized learning improved compared to an AI trained through self-play.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告ゲーム情報学（GI）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2026-02-23","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"19","bibliographicVolumeNumber":"2026-GI-57"}]},"relation_version_is_last":true,"weko_creator_id":"80578"},"id":2007486,"updated":"2026-02-17T06:52:12.064189+00:00","links":{}}