{"links":{},"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00234557","sets":["1164:1579:11464:11617"]},"path":["11617"],"owner":"44499","recid":"234557","title":["アクティベーション最大化に基づくAIアクセラレータの故障分類と予測"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-06-03"},"_buckets":{"deposit":"77d289c4-d3e8-4f6e-9224-ddca20f66a88"},"_deposit":{"id":"234557","pid":{"type":"depid","value":"234557","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"アクティベーション最大化に基づくAIアクセラレータの故障分類と予測","author_link":["639304","639301","639303","639302"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"アクティベーション最大化に基づくAIアクセラレータの故障分類と予測"},{"subitem_title":"Fault Classification and Prediction of AI Accelerators Based on Activation Maximization","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"信頼性・品質管理","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2024-06-03","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"千葉大学大学院融合理工学府"},{"subitem_text_value":"千葉大学大学院工学研究院"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Graduate School of Science and Engineering, Chiba University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Engineering, Chiba 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/234557/files/IPSJ-ARC24257013.pdf","label":"IPSJ-ARC24257013.pdf"},"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-ARC24257013.pdf","filesize":[{"value":"1.6 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"0","billingrole":"16"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_login","version_id":"2aa4100c-0b47-4bc1-a5b7-c5512d247d95","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 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":"馬, 善謀"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"難波, 一輝"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Shanmou, Ma","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Kazuteru, Namba","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10096105","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":"2188-8574","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"本論文では，ニューラルネットワークの解釈性技術を基盤とした方法を提案し，AI アクセラレータの故障を効果的に分類し予測する．AI アクセラレータの PE が故障する場合，モデル，入力，故障位置など，異なる要因により，故障がモデルの出力に与える影響も異なる．本論文で提案する手法は，ニューラルネットワークの解釈性，特にアクティベーション最大化技術を利用して，AI アクセラレータの故障を AI アクセラレータの性能に与える影響の程度に応じて，重要故障と非重要故障に分類するものである．提案手法 (AMM) は，故障のリアルタイム予測と分類において，モデルの解釈性を最大限に活用し，従来手法 (FTM) と比較して，故障カバレッジ率がわずかに 0.6% 低下する一方で，故障の分類時間は 61.5% 短縮された．また，故障位置の正確な特定を避け，モデル間の移行を容易にすることが可能となる．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"This research proposes a method based on neural network interpretability techniques to effectively classify and predict AI accelerator failures. When an AI accelerator's Processing Element (PE) fails, different factors such as the model, input, and location of the failure can affect the output of the model in various ways. This method utilizes neural network interpretability, particularly activation maximization techniques, to classify AI accelerator failures into critical and non-critical based on the extent of their impact on the accelerator's performance. This approach maximizes the interpretability of the model in real-time failure prediction and classification, and compared to traditional methods(FTM), the proposed method(AMM) has only a slight 0.6% decrease in fault coverage rate while reducing the classification time by 61.5%. It also avoids precise identification of the fault location and facilitates transitions between models.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告システム・アーキテクチャ（ARC）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2024-06-03","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"13","bibliographicVolumeNumber":"2024-ARC-257"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"created":"2025-01-19T01:36:20.938978+00:00","updated":"2025-01-19T09:46:15.572568+00:00","id":234557}