{"id":228568,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00228568","sets":["1164:10193:11168:11346"]},"path":["11346"],"owner":"44499","recid":"228568","title":["Factorization Machine with Annealing向けモデル学習手法の構築"],"pubdate":{"attribute_name":"公開日","attribute_value":"2023-10-19"},"_buckets":{"deposit":"36389991-cbb1-46ab-bd5e-6ed83d9209f6"},"_deposit":{"id":"228568","pid":{"type":"depid","value":"228568","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Factorization Machine with Annealing向けモデル学習手法の構築","author_link":["610214","610211","610213","610212"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Factorization Machine with Annealing向けモデル学習手法の構築"}]},"item_type_id":"4","publish_date":"2023-10-19","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"慶應義塾大学理工学部物理情報工学科"},{"subitem_text_value":"慶應義塾大学大学院理工学研究科基礎理工学専攻"},{"subitem_text_value":"慶應義塾大学大学院理工学研究科基礎理工学専攻"},{"subitem_text_value":"慶應義塾大学理工学部物理情報工学科／慶應義塾大学大学院理工学研究科基礎理工学専攻／慶應義塾大学ヒト生物学―微生物叢―量子計算研究センター(WPI-Bio2Q)／慶應義塾大学量子コンピューティングセンター／早稲田大学グリーン・コンピューティング・システム研究機構／東京工業大学国際先駆研究機構"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Department of Applied Physics and Physico-Informatics, Keio University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Science and Technology, Keio University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Science and Technology, Keio University","subitem_text_language":"en"},{"subitem_text_value":"Department of Applied Physics and Physico-Informatics, Keio University / Graduate School of Science and Technology, Keio University / Human Biology-Microbiome-Quantum Research Center (WPI-Bio2Q), Keio University / Quantum Computing Center, Keio University / Green Computing System Research Organization, Waseda University / International Research Frontiers Initiative, Tokyo Institute of Technology","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/228568/files/IPSJ-QS23010020.pdf","label":"IPSJ-QS23010020.pdf"},"date":[{"dateType":"Available","dateValue":"2025-10-19"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-QS23010020.pdf","filesize":[{"value":"1.0 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"53"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"c0b67bf1-4f67-4c6c-9c8b-087bcde9bc45","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2023 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"中野, 檀"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"山下, 将司"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"関, 優也"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"田中, 宗"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA12894105","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":"2435-6492","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"最適化すべき目的関数が陽に与えられておらず，シミュレーションや実験を通じて入力に対する出力結果が得られる場合の最適化問題を Black-Box 最適化（BB 最適化）問題と呼び，様々な場面で現れる困難な問題である．BB 最適化に対する新たな手法として提案された Factorization Machine with Annealing (FMA) [1] は，BB 関数を機械学習モデル Factorization Machine（FM）によって推定することで，それを用いてイジングマシン（Annealing machine）によって解を推論するという方法である．これら一連の操作を繰り返し行うことによって，BB 関数の呼び出し回数および計算量を抑えながら高精度な解探索ができると期待されている．しかし，問題によっては最適解探索が困難になる．本研究では，最適解探索が困難な BB 関数に対しても，従来の FMA に比べてより最適化性能の高い FMA を実現するために，新たな FM の学習手法を提案する．従来の学習手法では，FM で BB 関数の推定をする際に，その時点で存在する全データを用いて学習を行っていた．一方，本研究で提案する学習手法では，最新の Dlatest 個のデータのみを用いて BB 関数の推定を行った．巡回セールスマン問題に対して，提案手法を実装した FMA による BB 最適化を行った場合，特定の Dlatest において従来手法よりも最適化性能が向上することが示唆された．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"7","bibliographic_titles":[{"bibliographic_title":"研究報告量子ソフトウェア（QS）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2023-10-19","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"20","bibliographicVolumeNumber":"2023-QS-10"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"updated":"2025-01-19T11:48:27.543871+00:00","created":"2025-01-19T01:27:42.671010+00:00","links":{}}