{"created":"2025-01-19T01:43:31.921100+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00239727","sets":["934:989:11507:11752"]},"path":["11752"],"owner":"44499","recid":"239727","title":["Bスプライン関数を用いた演算ノード柔軟化による学習済みニューラルネットの精度向上手法"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-09-25"},"_buckets":{"deposit":"043a3f96-6840-4537-ae4a-7e15c0e703c9"},"_deposit":{"id":"239727","pid":{"type":"depid","value":"239727","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Bスプライン関数を用いた演算ノード柔軟化による学習済みニューラルネットの精度向上手法","author_link":["657223","657222","657225","657224"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Bスプライン関数を用いた演算ノード柔軟化による学習済みニューラルネットの精度向上手法"},{"subitem_title":"Improving Accuracy of Pre-trained Neural Networks by Making Computation Nodes More Flexible by B-spline Functions","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"[オリジナル論文] Bスプライン，ニューラルネットワーク，ファインチューニング，高精度化，軽量化","subitem_subject_scheme":"Other"}]},"item_type_id":"3","publish_date":"2024-09-25","item_3_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"横浜国立大学大学院環境情報学府"},{"subitem_text_value":"横浜国立大学大学院環境情報研究院"}]},"item_3_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Graduate School of Environment and Information Sciences, Yokohama National University","subitem_text_language":"en"},{"subitem_text_value":"Faculty of Environment and Information Sciences, Yokohama National 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/239727/files/IPSJ-TOM1703003.pdf","label":"IPSJ-TOM1703003.pdf"},"date":[{"dateType":"Available","dateValue":"2026-09-25"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-TOM1703003.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":"17"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"d6690dec-6ac4-4c67-80f5-0d51d5eef050","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_3_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"葛谷, 直規"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"長尾, 智晴"}],"nameIdentifiers":[{}]}]},"item_3_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Naoki, Kuzuya","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Tomoharu, Nagao","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_3_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA11464803","subitem_source_identifier_type":"NCID"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_6501","resourcetype":"journal article"}]},"item_3_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"1882-7780","subitem_source_identifier_type":"ISSN"}]},"item_3_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年，様々な分野でニューラルネットが用いられるようになっている．一方で，ニューラルネットの学習には多大な時間と計算リソース，そして高いスキルが必要となり，開発者の負担になっている．そのため，あらかじめ大規模データセットで学習された学習済みモデルの利用が広がっているが，それらのモデルが所望の精度を満たすとは限らない．そのような場合，さらに精度の良いモデルを探索する必要があり，煩雑で時間がかかる．そこで，一般的な学習済みネットワークのアーキテクチャはそのままに，演算ノードをより柔軟ではあるが機能的には等価なノードに置換し，再学習させることで精度向上を実現させる手法を提案する．具体的には，ネットワーク中の線形変換やRectified Linear Unit（ReLU）活性化関数を，それらの関数の上位集合であるBスプライン関数で置き換えて，再学習を行うことで精度向上させる手法を提案する．いくつかの分類問題で明らかな精度向上を確認した．","subitem_description_type":"Other"}]},"item_3_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"In recent years, the use of neural networks has become popular in many fields. On the other hand, training neural networks requires a large amount of time, computational resources, and high skills, which places a burden on developers. For this reason, the use of pre-trained models trained in advance on large-scale datasets are widely used. However, the problem is that these models do not always meet the desired accuracy. In such cases, it is necessary to search for a better model, which is cumbersome and time-consuming. Therefore, we propose a method to improve accuracy by replacing the computation nodes of a pre-trained network with more flexible, but functionally equivalent nodes and re-training them, while maintaining the original network architecture. Specifically, we propose a method to improve accuracy by replacing linear transformations and Rectified Linear Unit (ReLU) activation functions with B-spline functions, which are a superset of these functions, and re-training them. We confirmed a significant improvement in accuracy for some classification tasks.","subitem_description_type":"Other"}]},"item_3_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"24","bibliographic_titles":[{"bibliographic_title":"情報処理学会論文誌数理モデル化と応用（TOM）"}],"bibliographicPageStart":"13","bibliographicIssueDates":{"bibliographicIssueDate":"2024-09-25","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"3","bibliographicVolumeNumber":"17"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":239727,"updated":"2025-01-19T08:11:32.715731+00:00","links":{}}