Maximal Domain Independent Representations Improve Transfer Learning

The most effective domain adaptation (DA) involves the decomposition of data representation into a domain independent representation (DIRep), and a domain dependent representation (DDRep). A classifier is trained by using the DIRep of the labeled source images. Since the DIRep is domain invariant, t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Adrian Shuai, Bertino, Elisa, Dang, Xuan-Hong, Singla, Ankush, Tu, Yuhai, Wegman, Mark N
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Li, Adrian Shuai
Bertino, Elisa
Dang, Xuan-Hong
Singla, Ankush
Tu, Yuhai
Wegman, Mark N
description The most effective domain adaptation (DA) involves the decomposition of data representation into a domain independent representation (DIRep), and a domain dependent representation (DDRep). A classifier is trained by using the DIRep of the labeled source images. Since the DIRep is domain invariant, the classifier can be "transferred" to make predictions for the target domain with no (or few) labels. However, information useful for classification in the target domain can "hide" in the DDRep in current DA algorithms such as Domain-Separation-Networks (DSN). DSN's weak constraint to enforce orthogonality of DIRep and DDRep, allows this hiding and can result in poor performance. To address this shortcoming, we developed a new algorithm wherein a stronger constraint is imposed to minimize the DDRep by using a KL divergent loss for the DDRep in order to create the maximal DIRep that enhances transfer learning performance. By using synthetic data sets, we show explicitly that depending on initialization DSN with its weaker constraint can lead to sub-optimal solutions with poorer DA performance whereas our algorithm with maximal DIRep is robust against such perturbations. We demonstrate the equal-or-better performance of our approach against state-of-the-art algorithms by using several standard benchmark image datasets including Office. We further highlight the compatibility of our algorithm with pretrained models, extending its applicability and versatility in real-world scenarios.
doi_str_mv 10.48550/arxiv.2306.00262
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2306_00262</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2306_00262</sourcerecordid><originalsourceid>FETCH-LOGICAL-a672-50f55b22afbf27fe4f59af8831a1d6b45790599299b56d2a352ef995c223cfad3</originalsourceid><addsrcrecordid>eNotj81ugzAQhH3poUr7AD3VLwA1a9bgY5X0B4moUsUdLWG3QgoGmShK37407WVmTp_mU-ohM2leIponipfhnII1LjUGHNyq3Z4uw0hHvZtGGoKuQs8zrxFO-pPnyMu66DRMYdHVOMfpzLqJFBbhqGumGIbwdaduhI4L3__3RjWvL832Pak_3qrtc52QKyBBI4gdAEknUAjngp6kLG1GWe-6HAtv0HvwvkPXA1kEFu_xAGAPQr3dqMc_7FWjneN6PH63vzrtVcf-AKbuRbg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Maximal Domain Independent Representations Improve Transfer Learning</title><source>arXiv.org</source><creator>Li, Adrian Shuai ; Bertino, Elisa ; Dang, Xuan-Hong ; Singla, Ankush ; Tu, Yuhai ; Wegman, Mark N</creator><creatorcontrib>Li, Adrian Shuai ; Bertino, Elisa ; Dang, Xuan-Hong ; Singla, Ankush ; Tu, Yuhai ; Wegman, Mark N</creatorcontrib><description>The most effective domain adaptation (DA) involves the decomposition of data representation into a domain independent representation (DIRep), and a domain dependent representation (DDRep). A classifier is trained by using the DIRep of the labeled source images. Since the DIRep is domain invariant, the classifier can be "transferred" to make predictions for the target domain with no (or few) labels. However, information useful for classification in the target domain can "hide" in the DDRep in current DA algorithms such as Domain-Separation-Networks (DSN). DSN's weak constraint to enforce orthogonality of DIRep and DDRep, allows this hiding and can result in poor performance. To address this shortcoming, we developed a new algorithm wherein a stronger constraint is imposed to minimize the DDRep by using a KL divergent loss for the DDRep in order to create the maximal DIRep that enhances transfer learning performance. By using synthetic data sets, we show explicitly that depending on initialization DSN with its weaker constraint can lead to sub-optimal solutions with poorer DA performance whereas our algorithm with maximal DIRep is robust against such perturbations. We demonstrate the equal-or-better performance of our approach against state-of-the-art algorithms by using several standard benchmark image datasets including Office. We further highlight the compatibility of our algorithm with pretrained models, extending its applicability and versatility in real-world scenarios.</description><identifier>DOI: 10.48550/arxiv.2306.00262</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Learning</subject><creationdate>2023-05</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,777,882</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2306.00262$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2306.00262$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Li, Adrian Shuai</creatorcontrib><creatorcontrib>Bertino, Elisa</creatorcontrib><creatorcontrib>Dang, Xuan-Hong</creatorcontrib><creatorcontrib>Singla, Ankush</creatorcontrib><creatorcontrib>Tu, Yuhai</creatorcontrib><creatorcontrib>Wegman, Mark N</creatorcontrib><title>Maximal Domain Independent Representations Improve Transfer Learning</title><description>The most effective domain adaptation (DA) involves the decomposition of data representation into a domain independent representation (DIRep), and a domain dependent representation (DDRep). A classifier is trained by using the DIRep of the labeled source images. Since the DIRep is domain invariant, the classifier can be "transferred" to make predictions for the target domain with no (or few) labels. However, information useful for classification in the target domain can "hide" in the DDRep in current DA algorithms such as Domain-Separation-Networks (DSN). DSN's weak constraint to enforce orthogonality of DIRep and DDRep, allows this hiding and can result in poor performance. To address this shortcoming, we developed a new algorithm wherein a stronger constraint is imposed to minimize the DDRep by using a KL divergent loss for the DDRep in order to create the maximal DIRep that enhances transfer learning performance. By using synthetic data sets, we show explicitly that depending on initialization DSN with its weaker constraint can lead to sub-optimal solutions with poorer DA performance whereas our algorithm with maximal DIRep is robust against such perturbations. We demonstrate the equal-or-better performance of our approach against state-of-the-art algorithms by using several standard benchmark image datasets including Office. We further highlight the compatibility of our algorithm with pretrained models, extending its applicability and versatility in real-world scenarios.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj81ugzAQhH3poUr7AD3VLwA1a9bgY5X0B4moUsUdLWG3QgoGmShK37407WVmTp_mU-ohM2leIponipfhnII1LjUGHNyq3Z4uw0hHvZtGGoKuQs8zrxFO-pPnyMu66DRMYdHVOMfpzLqJFBbhqGumGIbwdaduhI4L3__3RjWvL832Pak_3qrtc52QKyBBI4gdAEknUAjngp6kLG1GWe-6HAtv0HvwvkPXA1kEFu_xAGAPQr3dqMc_7FWjneN6PH63vzrtVcf-AKbuRbg</recordid><startdate>20230531</startdate><enddate>20230531</enddate><creator>Li, Adrian Shuai</creator><creator>Bertino, Elisa</creator><creator>Dang, Xuan-Hong</creator><creator>Singla, Ankush</creator><creator>Tu, Yuhai</creator><creator>Wegman, Mark N</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20230531</creationdate><title>Maximal Domain Independent Representations Improve Transfer Learning</title><author>Li, Adrian Shuai ; Bertino, Elisa ; Dang, Xuan-Hong ; Singla, Ankush ; Tu, Yuhai ; Wegman, Mark N</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a672-50f55b22afbf27fe4f59af8831a1d6b45790599299b56d2a352ef995c223cfad3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Li, Adrian Shuai</creatorcontrib><creatorcontrib>Bertino, Elisa</creatorcontrib><creatorcontrib>Dang, Xuan-Hong</creatorcontrib><creatorcontrib>Singla, Ankush</creatorcontrib><creatorcontrib>Tu, Yuhai</creatorcontrib><creatorcontrib>Wegman, Mark N</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Li, Adrian Shuai</au><au>Bertino, Elisa</au><au>Dang, Xuan-Hong</au><au>Singla, Ankush</au><au>Tu, Yuhai</au><au>Wegman, Mark N</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Maximal Domain Independent Representations Improve Transfer Learning</atitle><date>2023-05-31</date><risdate>2023</risdate><abstract>The most effective domain adaptation (DA) involves the decomposition of data representation into a domain independent representation (DIRep), and a domain dependent representation (DDRep). A classifier is trained by using the DIRep of the labeled source images. Since the DIRep is domain invariant, the classifier can be "transferred" to make predictions for the target domain with no (or few) labels. However, information useful for classification in the target domain can "hide" in the DDRep in current DA algorithms such as Domain-Separation-Networks (DSN). DSN's weak constraint to enforce orthogonality of DIRep and DDRep, allows this hiding and can result in poor performance. To address this shortcoming, we developed a new algorithm wherein a stronger constraint is imposed to minimize the DDRep by using a KL divergent loss for the DDRep in order to create the maximal DIRep that enhances transfer learning performance. By using synthetic data sets, we show explicitly that depending on initialization DSN with its weaker constraint can lead to sub-optimal solutions with poorer DA performance whereas our algorithm with maximal DIRep is robust against such perturbations. We demonstrate the equal-or-better performance of our approach against state-of-the-art algorithms by using several standard benchmark image datasets including Office. We further highlight the compatibility of our algorithm with pretrained models, extending its applicability and versatility in real-world scenarios.</abstract><doi>10.48550/arxiv.2306.00262</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2306.00262
ispartof
issn
language eng
recordid cdi_arxiv_primary_2306_00262
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Learning
title Maximal Domain Independent Representations Improve Transfer Learning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-21T01%3A53%3A59IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Maximal%20Domain%20Independent%20Representations%20Improve%20Transfer%20Learning&rft.au=Li,%20Adrian%20Shuai&rft.date=2023-05-31&rft_id=info:doi/10.48550/arxiv.2306.00262&rft_dat=%3Carxiv_GOX%3E2306_00262%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true