Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data

The scarcity of class-labeled data is a ubiquitous bottleneck in many machine learning problems. While abundant unlabeled data typically exist and provide a potential solution, it is highly challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) clas...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yu, Bing, Sun, Ke, Wang, He, Lin, Zhouchen, Zhu, Zhanxing
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 Yu, Bing
Sun, Ke
Wang, He
Lin, Zhouchen
Zhu, Zhanxing
description The scarcity of class-labeled data is a ubiquitous bottleneck in many machine learning problems. While abundant unlabeled data typically exist and provide a potential solution, it is highly challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) classification and the conditional generation with extra unlabeled data \emph{simultaneously}. In particular, we present a novel training framework to jointly target both PU classification and conditional generation when exposed to extra data, especially out-of-distribution unlabeled data, by exploring the interplay between them: 1) enhancing the performance of PU classifiers with the assistance of a novel Classifier-Noise-Invariant Conditional GAN~(CNI-CGAN) that is robust to noisy labels, 2) leveraging extra data with predicted labels from a PU classifier to help the generation. Theoretically, we prove the optimal condition of CNI-CGAN, and experimentally, we conducted extensive evaluations on diverse datasets, verifying the simultaneous improvements in both classification and generation.
doi_str_mv 10.48550/arxiv.2006.07841
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2006_07841</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2006_07841</sourcerecordid><originalsourceid>FETCH-LOGICAL-a671-5deeaa380eecc5d8164b63a3f9d8bc90922607598b747b553a1cd377150b972e3</originalsourceid><addsrcrecordid>eNo1kMFOhDAURdm4MKMf4Mr-ANhSSsGdwXE0IdHouCav7UObdMqkdEb4e0fU1c1Nbk5yT5JcMZoVlRD0BsJkj1lOaZlRWRXsPJkaB-No-5mAN2SDHgNEJK-o7T4MGpybb8mb3R1cBI_DYSQvw2ijPWL67h0odA4NaRGCt_5jgTSDN6fF4MH9A0-FfNn4SdZTDEDuIcJFctaDG_HyL1fJ9mG9bR7T9nnz1Ny1KZSSpcIgAvCKImotTMXKQpUceF-bSuma1nleUinqSslCKiE4MG24lExQVcsc-Sq5_sUu17t9sDsIc_ejoFsU8G-Raliv</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data</title><source>arXiv.org</source><creator>Yu, Bing ; Sun, Ke ; Wang, He ; Lin, Zhouchen ; Zhu, Zhanxing</creator><creatorcontrib>Yu, Bing ; Sun, Ke ; Wang, He ; Lin, Zhouchen ; Zhu, Zhanxing</creatorcontrib><description>The scarcity of class-labeled data is a ubiquitous bottleneck in many machine learning problems. While abundant unlabeled data typically exist and provide a potential solution, it is highly challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) classification and the conditional generation with extra unlabeled data \emph{simultaneously}. In particular, we present a novel training framework to jointly target both PU classification and conditional generation when exposed to extra data, especially out-of-distribution unlabeled data, by exploring the interplay between them: 1) enhancing the performance of PU classifiers with the assistance of a novel Classifier-Noise-Invariant Conditional GAN~(CNI-CGAN) that is robust to noisy labels, 2) leveraging extra data with predicted labels from a PU classifier to help the generation. Theoretically, we prove the optimal condition of CNI-CGAN, and experimentally, we conducted extensive evaluations on diverse datasets, verifying the simultaneous improvements in both classification and generation.</description><identifier>DOI: 10.48550/arxiv.2006.07841</identifier><language>eng</language><subject>Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2020-06</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</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/2006.07841$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2006.07841$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Yu, Bing</creatorcontrib><creatorcontrib>Sun, Ke</creatorcontrib><creatorcontrib>Wang, He</creatorcontrib><creatorcontrib>Lin, Zhouchen</creatorcontrib><creatorcontrib>Zhu, Zhanxing</creatorcontrib><title>Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data</title><description>The scarcity of class-labeled data is a ubiquitous bottleneck in many machine learning problems. While abundant unlabeled data typically exist and provide a potential solution, it is highly challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) classification and the conditional generation with extra unlabeled data \emph{simultaneously}. In particular, we present a novel training framework to jointly target both PU classification and conditional generation when exposed to extra data, especially out-of-distribution unlabeled data, by exploring the interplay between them: 1) enhancing the performance of PU classifiers with the assistance of a novel Classifier-Noise-Invariant Conditional GAN~(CNI-CGAN) that is robust to noisy labels, 2) leveraging extra data with predicted labels from a PU classifier to help the generation. Theoretically, we prove the optimal condition of CNI-CGAN, and experimentally, we conducted extensive evaluations on diverse datasets, verifying the simultaneous improvements in both classification and generation.</description><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNo1kMFOhDAURdm4MKMf4Mr-ANhSSsGdwXE0IdHouCav7UObdMqkdEb4e0fU1c1Nbk5yT5JcMZoVlRD0BsJkj1lOaZlRWRXsPJkaB-No-5mAN2SDHgNEJK-o7T4MGpybb8mb3R1cBI_DYSQvw2ijPWL67h0odA4NaRGCt_5jgTSDN6fF4MH9A0-FfNn4SdZTDEDuIcJFctaDG_HyL1fJ9mG9bR7T9nnz1Ny1KZSSpcIgAvCKImotTMXKQpUceF-bSuma1nleUinqSslCKiE4MG24lExQVcsc-Sq5_sUu17t9sDsIc_ejoFsU8G-Raliv</recordid><startdate>20200614</startdate><enddate>20200614</enddate><creator>Yu, Bing</creator><creator>Sun, Ke</creator><creator>Wang, He</creator><creator>Lin, Zhouchen</creator><creator>Zhu, Zhanxing</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20200614</creationdate><title>Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data</title><author>Yu, Bing ; Sun, Ke ; Wang, He ; Lin, Zhouchen ; Zhu, Zhanxing</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a671-5deeaa380eecc5d8164b63a3f9d8bc90922607598b747b553a1cd377150b972e3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Yu, Bing</creatorcontrib><creatorcontrib>Sun, Ke</creatorcontrib><creatorcontrib>Wang, He</creatorcontrib><creatorcontrib>Lin, Zhouchen</creatorcontrib><creatorcontrib>Zhu, Zhanxing</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Yu, Bing</au><au>Sun, Ke</au><au>Wang, He</au><au>Lin, Zhouchen</au><au>Zhu, Zhanxing</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data</atitle><date>2020-06-14</date><risdate>2020</risdate><abstract>The scarcity of class-labeled data is a ubiquitous bottleneck in many machine learning problems. While abundant unlabeled data typically exist and provide a potential solution, it is highly challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) classification and the conditional generation with extra unlabeled data \emph{simultaneously}. In particular, we present a novel training framework to jointly target both PU classification and conditional generation when exposed to extra data, especially out-of-distribution unlabeled data, by exploring the interplay between them: 1) enhancing the performance of PU classifiers with the assistance of a novel Classifier-Noise-Invariant Conditional GAN~(CNI-CGAN) that is robust to noisy labels, 2) leveraging extra data with predicted labels from a PU classifier to help the generation. Theoretically, we prove the optimal condition of CNI-CGAN, and experimentally, we conducted extensive evaluations on diverse datasets, verifying the simultaneous improvements in both classification and generation.</abstract><doi>10.48550/arxiv.2006.07841</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2006.07841
ispartof
issn
language eng
recordid cdi_arxiv_primary_2006_07841
source arXiv.org
subjects Computer Science - Learning
Statistics - Machine Learning
title Classify and Generate Reciprocally: Simultaneous Positive-Unlabelled Learning and Conditional Generation with Extra Data
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-18T15%3A59%3A02IST&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=Classify%20and%20Generate%20Reciprocally:%20Simultaneous%20Positive-Unlabelled%20Learning%20and%20Conditional%20Generation%20with%20Extra%20Data&rft.au=Yu,%20Bing&rft.date=2020-06-14&rft_id=info:doi/10.48550/arxiv.2006.07841&rft_dat=%3Carxiv_GOX%3E2006_07841%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