{"id":211911,"updated":"2025-01-19T17:38:07.532647+00:00","links":{},"created":"2025-01-19T01:13:01.732214+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00211911","sets":["1164:1579:10482:10619"]},"path":["10619"],"owner":"44499","recid":"211911","title":["DPDKを用いた分散深層強化学習における経験サンプリングの高速化"],"pubdate":{"attribute_name":"公開日","attribute_value":"2021-07-13"},"_buckets":{"deposit":"a39f1e60-9717-4658-82cc-e0f222b4b32d"},"_deposit":{"id":"211911","pid":{"type":"depid","value":"211911","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"DPDKを用いた分散深層強化学習における経験サンプリングの高速化","author_link":["539387","539386","539388","539385"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"DPDKを用いた分散深層強化学習における経験サンプリングの高速化"},{"subitem_title":"A DPDK-Based Acceleration Method for Experience Sampling of Distributed Reinforcement Learning","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"分散コンピューティング","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2021-07-13","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"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":"Graduate School 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"}]},"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/211911/files/IPSJ-ARC21245013.pdf","label":"IPSJ-ARC21245013.pdf"},"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-ARC21245013.pdf","filesize":[{"value":"2.1 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"0","billingrole":"16"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_login","version_id":"5348fba2-55f1-4e59-9d35-70785358922b","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":"古川, 雅輝"}],"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":"Masaki, Furukawa","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":"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":"DQN (Deep Q-Network) に代表される深層強化学習の性能を向上させるため，分散深層強化学習では，複数の計算機をネットワーク接続した計算機クラスタが用いられる．計算機クラスタを用いた分散深層強化学習では，環境空間の探索により経験を獲得する Actor と深層学習モデルを最適化する Learner の間で，経験サイズや Actor 数に応じたデータ転送が頻繁に発生するため，通信コストが分散学習の性能向上を妨げる．そこで，本研究では 40GbE (40Gbit Ethernet) ネットワークで接続された Actor と Learner の間に，DPDK によって低遅延化されたインメモリデータベースや経験再生メモリを導入することで，分散深層強化学習における通信コストの削減を図る．DPDK を用いたカーネルバイパスによるネットワーク最適化によって，共有メモリへのアクセス遅延は 32.7%～58.9% 削減された．また，DPDK  ベースの優先度付き経験再生メモリをネットワーク上に実装することで，経験再生メモリへのアクセス遅延は 11.7%～28.1% 改善し，優先度付き経験サンプリングにおける通信遅延は 21.9%～29.1% 削減された．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告システム・アーキテクチャ（ARC）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2021-07-13","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"13","bibliographicVolumeNumber":"2021-ARC-245"}]},"relation_version_is_last":true,"weko_creator_id":"44499"}}