Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation

Domain adaptation is an approach to transfer knowledge from a certain source domain to another target domain. Several studies are recently reported on partial domain adaptation (PDA), where the class set of the source domain is larger than that of the target domain as a more realistic setting, but t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Shisutemu Seigyo Jouhou Gakkai rombunshi Control and Information Engineers, 2022/05/15, Vol.35(5), pp.101-108
Hauptverfasser: Kono, Seita, Ueda, Takaya, Takano, Ryo, Nishikawa, Ikuko
Format: Artikel
Sprache:eng ; jpn
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 108
container_issue 5
container_start_page 101
container_title Shisutemu Seigyo Jouhou Gakkai rombunshi
container_volume 35
creator Kono, Seita
Ueda, Takaya
Takano, Ryo
Nishikawa, Ikuko
description Domain adaptation is an approach to transfer knowledge from a certain source domain to another target domain. Several studies are recently reported on partial domain adaptation (PDA), where the class set of the source domain is larger than that of the target domain as a more realistic setting, but then the source domain specific classes make the adaptation more difficult. Most existing methods for PDA give small weights to the source domain specific classes to prevent the target data from being matched. The present paper proposes a PDA method which introduces a novel mechanism that gives additional weights to an individual target data by estimating the probability that the data belongs to each source class. The estimation is given by multiple discriminators that measure the distance between the data distribution of each source class and the entire target data distribution through adversarial training against a data encoder. Computer experiments using two handwritten digit datasets as two domains show that the proposed method achieves more stable and accurate domain adaptation compared with state-of-the-art existing methods for PDA.
doi_str_mv 10.5687/iscie.35.101
format Article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_journals_2709084463</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2709084463</sourcerecordid><originalsourceid>FETCH-LOGICAL-c1213-9295717699e033d47382c950475fd94c167bff31f0c89a81267bb2718ebd6f413</originalsourceid><addsrcrecordid>eNpFkE9LAzEQxYMoWGpvfoAFr27NbDab5CSl1j8g2EMF8RKy2aRN2W5qshX67d1tSz0N8-Y3b3iD0C3gMS04e3BROzMmdAwYLtAgA05TDvB1iQZA8iylRcGv0ShGV2ICLAcgdIC-Z7F1G9U63yTeJtNaxZjMgy9V6WrX7pN2FfxuuUom1a8JUQWn6mQRlGtcs0ysD8lchbYXn_ymUztObduD3w26sqqOZnSqQ_T5PFtMX9P3j5e36eQ91ZABSUUmKANWCGEwIVXOCM-0oDhn1FYi11Cw0loCFmsuFIes68uMATdlVdgcyBDdHX23wf_sTGzl2u9C052UGcMC8zwvSEfdHykdfIzBWLkNXfCwl4Bl_0B5eKAktBN608cjvo6tWpoz3IfVtfmH6WnjPNErFaRpyB_BT3rK</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2709084463</pqid></control><display><type>article</type><title>Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation</title><source>J-STAGE (Japan Science &amp; Technology Information Aggregator, Electronic) Freely Available Titles - Japanese</source><source>EZB-FREE-00999 freely available EZB journals</source><creator>Kono, Seita ; Ueda, Takaya ; Takano, Ryo ; Nishikawa, Ikuko</creator><creatorcontrib>Kono, Seita ; Ueda, Takaya ; Takano, Ryo ; Nishikawa, Ikuko</creatorcontrib><description>Domain adaptation is an approach to transfer knowledge from a certain source domain to another target domain. Several studies are recently reported on partial domain adaptation (PDA), where the class set of the source domain is larger than that of the target domain as a more realistic setting, but then the source domain specific classes make the adaptation more difficult. Most existing methods for PDA give small weights to the source domain specific classes to prevent the target data from being matched. The present paper proposes a PDA method which introduces a novel mechanism that gives additional weights to an individual target data by estimating the probability that the data belongs to each source class. The estimation is given by multiple discriminators that measure the distance between the data distribution of each source class and the entire target data distribution through adversarial training against a data encoder. Computer experiments using two handwritten digit datasets as two domains show that the proposed method achieves more stable and accurate domain adaptation compared with state-of-the-art existing methods for PDA.</description><identifier>ISSN: 1342-5668</identifier><identifier>EISSN: 2185-811X</identifier><identifier>DOI: 10.5687/iscie.35.101</identifier><language>eng ; jpn</language><publisher>Kyoto: THE INSTITUTE OF SYSTEMS, CONTROL AND INFORMATION ENGINEERS (ISCIE)</publisher><subject>Adaptation ; adversarial training ; Coders ; data distribution ; Discriminators ; Estimation ; Handwriting ; Knowledge management ; Kullback-Leibler divergence ; partial domain adaptation ; Personal digital assistants ; Training</subject><ispartof>Transactions of the Institute of Systems, Control and Information Engineers, 2022/05/15, Vol.35(5), pp.101-108</ispartof><rights>2022 The Institute of Systems, Control and Information Engineers</rights><rights>Copyright Japan Science and Technology Agency 2022</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c1213-9295717699e033d47382c950475fd94c167bff31f0c89a81267bb2718ebd6f413</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,780,784,1883,27924,27925</link.rule.ids></links><search><creatorcontrib>Kono, Seita</creatorcontrib><creatorcontrib>Ueda, Takaya</creatorcontrib><creatorcontrib>Takano, Ryo</creatorcontrib><creatorcontrib>Nishikawa, Ikuko</creatorcontrib><title>Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation</title><title>Shisutemu Seigyo Jouhou Gakkai rombunshi</title><addtitle>Transactions of the Institute of Systems, Control and Information Engineers</addtitle><description>Domain adaptation is an approach to transfer knowledge from a certain source domain to another target domain. Several studies are recently reported on partial domain adaptation (PDA), where the class set of the source domain is larger than that of the target domain as a more realistic setting, but then the source domain specific classes make the adaptation more difficult. Most existing methods for PDA give small weights to the source domain specific classes to prevent the target data from being matched. The present paper proposes a PDA method which introduces a novel mechanism that gives additional weights to an individual target data by estimating the probability that the data belongs to each source class. The estimation is given by multiple discriminators that measure the distance between the data distribution of each source class and the entire target data distribution through adversarial training against a data encoder. Computer experiments using two handwritten digit datasets as two domains show that the proposed method achieves more stable and accurate domain adaptation compared with state-of-the-art existing methods for PDA.</description><subject>Adaptation</subject><subject>adversarial training</subject><subject>Coders</subject><subject>data distribution</subject><subject>Discriminators</subject><subject>Estimation</subject><subject>Handwriting</subject><subject>Knowledge management</subject><subject>Kullback-Leibler divergence</subject><subject>partial domain adaptation</subject><subject>Personal digital assistants</subject><subject>Training</subject><issn>1342-5668</issn><issn>2185-811X</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><recordid>eNpFkE9LAzEQxYMoWGpvfoAFr27NbDab5CSl1j8g2EMF8RKy2aRN2W5qshX67d1tSz0N8-Y3b3iD0C3gMS04e3BROzMmdAwYLtAgA05TDvB1iQZA8iylRcGv0ShGV2ICLAcgdIC-Z7F1G9U63yTeJtNaxZjMgy9V6WrX7pN2FfxuuUom1a8JUQWn6mQRlGtcs0ysD8lchbYXn_ymUztObduD3w26sqqOZnSqQ_T5PFtMX9P3j5e36eQ91ZABSUUmKANWCGEwIVXOCM-0oDhn1FYi11Cw0loCFmsuFIes68uMATdlVdgcyBDdHX23wf_sTGzl2u9C052UGcMC8zwvSEfdHykdfIzBWLkNXfCwl4Bl_0B5eKAktBN608cjvo6tWpoz3IfVtfmH6WnjPNErFaRpyB_BT3rK</recordid><startdate>20220515</startdate><enddate>20220515</enddate><creator>Kono, Seita</creator><creator>Ueda, Takaya</creator><creator>Takano, Ryo</creator><creator>Nishikawa, Ikuko</creator><general>THE INSTITUTE OF SYSTEMS, CONTROL AND INFORMATION ENGINEERS (ISCIE)</general><general>Japan Science and Technology Agency</general><scope>AAYXX</scope><scope>CITATION</scope><scope>JQ2</scope></search><sort><creationdate>20220515</creationdate><title>Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation</title><author>Kono, Seita ; Ueda, Takaya ; Takano, Ryo ; Nishikawa, Ikuko</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c1213-9295717699e033d47382c950475fd94c167bff31f0c89a81267bb2718ebd6f413</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng ; jpn</language><creationdate>2022</creationdate><topic>Adaptation</topic><topic>adversarial training</topic><topic>Coders</topic><topic>data distribution</topic><topic>Discriminators</topic><topic>Estimation</topic><topic>Handwriting</topic><topic>Knowledge management</topic><topic>Kullback-Leibler divergence</topic><topic>partial domain adaptation</topic><topic>Personal digital assistants</topic><topic>Training</topic><toplevel>online_resources</toplevel><creatorcontrib>Kono, Seita</creatorcontrib><creatorcontrib>Ueda, Takaya</creatorcontrib><creatorcontrib>Takano, Ryo</creatorcontrib><creatorcontrib>Nishikawa, Ikuko</creatorcontrib><collection>CrossRef</collection><collection>ProQuest Computer Science Collection</collection><jtitle>Shisutemu Seigyo Jouhou Gakkai rombunshi</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Kono, Seita</au><au>Ueda, Takaya</au><au>Takano, Ryo</au><au>Nishikawa, Ikuko</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation</atitle><jtitle>Shisutemu Seigyo Jouhou Gakkai rombunshi</jtitle><addtitle>Transactions of the Institute of Systems, Control and Information Engineers</addtitle><date>2022-05-15</date><risdate>2022</risdate><volume>35</volume><issue>5</issue><spage>101</spage><epage>108</epage><pages>101-108</pages><issn>1342-5668</issn><eissn>2185-811X</eissn><abstract>Domain adaptation is an approach to transfer knowledge from a certain source domain to another target domain. Several studies are recently reported on partial domain adaptation (PDA), where the class set of the source domain is larger than that of the target domain as a more realistic setting, but then the source domain specific classes make the adaptation more difficult. Most existing methods for PDA give small weights to the source domain specific classes to prevent the target data from being matched. The present paper proposes a PDA method which introduces a novel mechanism that gives additional weights to an individual target data by estimating the probability that the data belongs to each source class. The estimation is given by multiple discriminators that measure the distance between the data distribution of each source class and the entire target data distribution through adversarial training against a data encoder. Computer experiments using two handwritten digit datasets as two domains show that the proposed method achieves more stable and accurate domain adaptation compared with state-of-the-art existing methods for PDA.</abstract><cop>Kyoto</cop><pub>THE INSTITUTE OF SYSTEMS, CONTROL AND INFORMATION ENGINEERS (ISCIE)</pub><doi>10.5687/iscie.35.101</doi><tpages>8</tpages><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 1342-5668
ispartof Transactions of the Institute of Systems, Control and Information Engineers, 2022/05/15, Vol.35(5), pp.101-108
issn 1342-5668
2185-811X
language eng ; jpn
recordid cdi_proquest_journals_2709084463
source J-STAGE (Japan Science & Technology Information Aggregator, Electronic) Freely Available Titles - Japanese; EZB-FREE-00999 freely available EZB journals
subjects Adaptation
adversarial training
Coders
data distribution
Discriminators
Estimation
Handwriting
Knowledge management
Kullback-Leibler divergence
partial domain adaptation
Personal digital assistants
Training
title Estimation of Class Probability through Adversarial Training for Partial Domain Adaptation
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T19%3A07%3A48IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Estimation%20of%20Class%20Probability%20through%20Adversarial%20Training%20for%20Partial%20Domain%20Adaptation&rft.jtitle=Shisutemu%20Seigyo%20Jouhou%20Gakkai%20rombunshi&rft.au=Kono,%20Seita&rft.date=2022-05-15&rft.volume=35&rft.issue=5&rft.spage=101&rft.epage=108&rft.pages=101-108&rft.issn=1342-5668&rft.eissn=2185-811X&rft_id=info:doi/10.5687/iscie.35.101&rft_dat=%3Cproquest_cross%3E2709084463%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2709084463&rft_id=info:pmid/&rfr_iscdi=true