{"created":"2025-01-19T01:11:45.244518+00:00","updated":"2025-01-19T18:07:26.632955+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00210538","sets":["1164:2822:10563:10564"]},"path":["10564"],"owner":"44499","recid":"210538","title":["Neural ODEを用いたエッジデバイス向けドメイン適応手法"],"pubdate":{"attribute_name":"公開日","attribute_value":"2021-03-18"},"_buckets":{"deposit":"f7b755e8-5ecc-4fa6-8544-97f687443a08"},"_deposit":{"id":"210538","pid":{"type":"depid","value":"210538","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Neural ODEを用いたエッジデバイス向けドメイン適応手法","author_link":["533292","533295","533293","533296","533294","533291"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Neural ODEを用いたエッジデバイス向けドメイン適応手法"},{"subitem_title":"A Neural ODE Based Domain Adaptation Method for Edge Devices","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"機械学習","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2021-03-18","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"},{"subitem_text_value":"慶應義塾大学大学院理工学研究科"},{"subitem_text_value":"慶應義塾大学理工学部／慶應義塾大学大学院理工学研究科 "}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Faculty 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":"Faculty of Science and Technology, Keio University / Graduate School of Science and Technology, Keio University ","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/210538/files/IPSJ-EMB21056040.pdf","label":"IPSJ-EMB21056040.pdf"},"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-EMB21056040.pdf","filesize":[{"value":"1.9 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"0","billingrole":"42"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_login","version_id":"3753e7e3-280c-4f5d-a09e-3c0eaee82c17","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2021 by the Institute of Electronics, Information and Communication Engineers This SIG report is only available to those in membership of the SIG."}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"川上, 大輝"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"渡邉, 寛悠"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"松谷, 宏紀"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Hiroki, Kawakami","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hirohisa, Watanabe","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hiroki, Matsutani","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA12149313","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":"2188-868X","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年，画像認識の分野では深層ニューラルネットワークの層を深くすることによって高い精度を実現している．しかし，エッジデバイスには計算リソースの制限があり，大規模なニューラルネットワークを用いることは難しい．また，深層学習の課題にドメインシフトがあり，これに適応させるドメイン適応という技術がある．これらの背景をもとに，本研究ではエッジデバイスとして小規模 FPGA（Field-Programmable Gate Array）を用いることを前提としたドメイン適応手法を提案する．モデルに小規模 FPGA に実装可能な Neural ODE を用いており，学習に蒸留を用いている．評価の 1 つとして，1 桁の数字が描かれたデータセットを用いたところ，教師モデルから 1 つ目の生徒モデルへの精度向上が 9.9%，1 つ目の生徒モデルから 2 つ目の生徒モデルへの精度向上が 1.4% となった．提案手法は，パラメータ数を削減することによって小規模 FPGA に一部オフロードを可能としつつ，ドメイン適応によって精度の向上が見られた．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告組込みシステム（EMB）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2021-03-18","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"40","bibliographicVolumeNumber":"2021-EMB-56"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":210538,"links":{}}