{"updated":"2025-01-19T10:49:53.441581+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00231263","sets":["1164:4179:11237:11430"]},"path":["11430"],"owner":"44499","recid":"231263","title":["ランダム巡回ベクトルを用いたマルチラベル学習"],"pubdate":{"attribute_name":"公開日","attribute_value":"2023-11-25"},"_buckets":{"deposit":"1258f7fe-98ff-4986-a632-3942026359cc"},"_deposit":{"id":"231263","pid":{"type":"depid","value":"231263","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"ランダム巡回ベクトルを用いたマルチラベル学習","author_link":["624072","624074","624075","624073"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"ランダム巡回ベクトルを用いたマルチラベル学習"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"モデルとデータ活用","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2023-11-25","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"北海道大学"},{"subitem_text_value":"北海道大学"},{"subitem_text_value":"北海道大学"},{"subitem_text_value":"奈良先端科学技術大学院大学"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Hokkaido University","subitem_text_language":"en"},{"subitem_text_value":"Hokkaido University","subitem_text_language":"en"},{"subitem_text_value":"Hokkaido University","subitem_text_language":"en"},{"subitem_text_value":"Nara Institute of Science and 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/231263/files/IPSJ-NL23258005.pdf","label":"IPSJ-NL23258005.pdf"},"date":[{"dateType":"Available","dateValue":"2025-11-25"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-NL23258005.pdf","filesize":[{"value":"2.7 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":"23"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"04503271-d383-4a3b-9834-60d20c2f7d92","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":"西田, 拳"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"林, 克彦"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"町, 光二郎"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"上垣外, 英剛"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10115061","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":"2188-8779","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"大規模マルチラベル学習（XMC; Extreme Multi-label Classification）は，数十万規模の膨大なラベル集合からテキストなどのデータ事例に割り当てられたラベル集合を予測するタスクである．近年，深層学習（DNN）が XMC タスクにおいて，高い性能を示している一方で，膨大な数のラベルを効率的に処理することは依然として大きな課題となっている．本研究では効率的な XMC 学習に向けて，複素偏角を要素とした巡回ベクトルを利用し，DNN の出力層と損失関数の計算効率を大幅に向上させる手法を提案する．提案手法では，データ事例に割り当てられたラベル集合を低次元の巡回ベクトルに符号化し，DNN の全結合層において，その巡回ベクトルを直接予測することで，出力層のサイズを大幅に削減する．人工データセットを用いた理論実験において，通常の実数値ベクトルと比較して，巡回ベクトルが優れたラベル符号化能力とラベル検索能力を持つことが確認された．さらに，実際の XMC データセットを用いた分類実験では，出力層のサイズを最大で 99% 削減しながら，高いタスク性能を達成することが示された．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"12","bibliographic_titles":[{"bibliographic_title":"研究報告自然言語処理（NL）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2023-11-25","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"5","bibliographicVolumeNumber":"2023-NL-258"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"created":"2025-01-19T01:31:27.608090+00:00","id":231263,"links":{}}