{"id":236257,"updated":"2025-01-19T09:19:51.440447+00:00","links":{},"created":"2025-01-19T01:38:12.565433+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00236257","sets":["6504:11678:11689"]},"path":["11689"],"owner":"44499","recid":"236257","title":["ダミーデータを用いた低次元特徴量空間のFew-shot学習性能の改善"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-03-01"},"_buckets":{"deposit":"41266a70-052a-4d10-96af-b0f4d5117d98"},"_deposit":{"id":"236257","pid":{"type":"depid","value":"236257","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"ダミーデータを用いた低次元特徴量空間のFew-shot学習性能の改善","author_link":["645749","645750"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"ダミーデータを用いた低次元特徴量空間のFew-shot学習性能の改善"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"人工知能と認知科学","subitem_subject_scheme":"Other"}]},"item_type_id":"22","publish_date":"2024-03-01","item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_22_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"法大"},{"subitem_text_value":"法大"}]},"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/236257/files/IPSJ-Z86-7V-05.pdf","label":"IPSJ-Z86-7V-05.pdf"},"date":[{"dateType":"Available","dateValue":"2024-07-03"}],"format":"application/pdf","filename":"IPSJ-Z86-7V-05.pdf","filesize":[{"value":"402.3 kB"}],"mimetype":"application/pdf","accessrole":"open_date","version_id":"b9c7741d-5b55-4add-a295-0abc9d6c4b1b","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_22_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"坂, 侑拓"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"藤田, 悟"}],"nameIdentifiers":[{}]}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_5794","resourcetype":"conference paper"}]},"item_22_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN00349328","subitem_source_identifier_type":"NCID"}]},"item_22_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"メタ学習の一部であるFew-shot学習では、学習時に存在しないクラスについて、少ないデータが与えられるため、テストクラスの予測が困難となる課題が存在する。本研究では、データを数値に写像する特徴空間の最適化を目指し、学習クラスに存在しないダミーデータの生成と活用法を提案する。具体的には、学習クラスの一部を取り出し、特徴空間上での各クラスの特徴を持つようなダミーデータを生成し、学習クラスとダミークラスが離れるように学習する。先行研究であるPrototypical Networksに倣い、提案されている損失関数にダミーデータとの距離を追加したモデルを作成した。一部のデータセットで、低次元の特徴空間という条件の下、先行研究の精度を上回る結果を得た。","subitem_description_type":"Other"}]},"item_22_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"850","bibliographic_titles":[{"bibliographic_title":"第86回全国大会講演論文集"}],"bibliographicPageStart":"849","bibliographicIssueDates":{"bibliographicIssueDate":"2024-03-01","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"1","bibliographicVolumeNumber":"2024"}]},"relation_version_is_last":true,"weko_creator_id":"44499"}}