{"id":78719,"updated":"2025-01-21T20:24:24.440343+00:00","links":{},"created":"2025-01-18T23:33:52.964352+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00078719","sets":["1164:2735:6337:6602"]},"path":["6602"],"owner":"10","recid":"78719","title":["頑健なスパースカーネル分類器の学習"],"pubdate":{"attribute_name":"公開日","attribute_value":"2011-11-24"},"_buckets":{"deposit":"eae75cca-872a-43ba-9fe9-37aa99dcf036"},"_deposit":{"id":"78719","pid":{"type":"depid","value":"78719","revision_id":0},"owners":[10],"status":"published","created_by":10},"item_title":"頑健なスパースカーネル分類器の学習","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"頑健なスパースカーネル分類器の学習"},{"subitem_title":"Learning Robust Sparse Kernel Classifiers","subitem_title_language":"en"}]},"item_type_id":"4","publish_date":"2011-11-24","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"神戸大学"},{"subitem_text_value":"神戸大学"},{"subitem_text_value":"神戸大学"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Kobe University","subitem_text_language":"en"},{"subitem_text_value":"Kobe University","subitem_text_language":"en"},{"subitem_text_value":"Kobe 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/78719/files/IPSJ-MPS11086002.pdf"},"date":[{"dateType":"Available","dateValue":"2013-11-24"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-MPS11086002.pdf","filesize":[{"value":"252.8 kB"}],"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":"caebde6f-8dcb-4d9c-b246-af49ac184e75","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2011 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"ブロンデルマチュー"},{"creatorName":"関, 和広"},{"creatorName":"上原, 邦昭"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Mathieu, Blondel","creatorNameLang":"en"},{"creatorName":"Kazuhiro, Seki","creatorNameLang":"en"},{"creatorName":"Kuniaki, Uehara","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10505667","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_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"カーネル分類器は多くのデータセットに対して優れた精度を示すことが分かっている．しかし，カーネル分類器のモデルの複雑性は訓練事例数に応じて線形に増加するため，訓練データの規模が大きくなるほど効果的にカーネル分類器を学習することが難しくなる．本研究では，スパースカーネル分類器を学習するための新しい逐次最適化アルゴリズムを提案する．提案アルゴリズムは，カーネルパーセプトロンと kernel matching pursuit に着想を得たものであり，a） 訓練データを有効に使用できる，b） ラベルノイズに頑健である，c） 任意の損失関数を利用できる，d） 実装も容易であるという多くの特長がある．複数のデータセットで評価実験を行ったところ，多くの実験設定において，提案手法は従来手法と同等か高い精度を示すことが明らかになった．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Despite state-of-the-art accuracy on many real-world datasets, kernel classifiers remain notoriously difficult to train efficiently because the model complexity has a linear dependency with the number of training instances. In this paper, we propose a novel incremental optimization algorithm for learning sparse kernel classifiers in the primal. Strongly influenced by the kernel perceptron and kernel matching pursuit, our algorithm makes efficient use of training data, is robust to label noise, can employ any convex subdifferentiable loss function and is simple to implement. Extensive experiments on several standard datasets show that our algorithm achieves comparable or better performance than several existing methods.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"8","bibliographic_titles":[{"bibliographic_title":"研究報告数理モデル化と問題解決（MPS）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2011-11-24","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"2","bibliographicVolumeNumber":"2011-MPS-86"}]},"relation_version_is_last":true,"weko_creator_id":"10"}}