Serveur d'exploration sur l'opéra

Attention, ce site est en cours de développement !
Attention, site généré par des moyens informatiques à partir de corpus bruts.
Les informations ne sont donc pas validées.

Online Pairwise Learning Algorithms.

Identifieur interne : 000026 ( Main/Merge ); précédent : 000025; suivant : 000027

Online Pairwise Learning Algorithms.

Auteurs : Yiming Ying [États-Unis] ; Ding-Xuan Zhou [Hong Kong]

Source :

RBID : pubmed:26890352

Abstract

Pairwise learning usually refers to a learning task that involves a loss function depending on pairs of examples, among which the most notable ones are bipartite ranking, metric learning, and AUC maximization. In this letter we study an online algorithm for pairwise learning with a least-square loss function in an unconstrained setting of a reproducing kernel Hilbert space (RKHS) that we refer to as the Online Pairwise lEaRning Algorithm (OPERA). In contrast to existing works (Kar, Sriperumbudur, Jain, & Karnick, 2013 ; Wang, Khardon, Pechyony, & Jones, 2012 ), which require that the iterates are restricted to a bounded domain or the loss function is strongly convex, OPERA is associated with a non-strongly convex objective function and learns the target function in an unconstrained RKHS. Specifically, we establish a general theorem that guarantees the almost sure convergence for the last iterate of OPERA without any assumptions on the underlying distribution. Explicit convergence rates are derived under the condition of polynomially decaying step sizes. We also establish an interesting property for a family of widely used kernels in the setting of pairwise learning and illustrate the convergence results using such kernels. Our methodology mainly depends on the characterization of RKHSs using its associated integral operators and probability inequalities for random variables with values in a Hilbert space.

DOI: 10.1162/NECO_a_00817
PubMed: 26890352

Links toward previous steps (curation, corpus...)


Links to Exploration step

pubmed:26890352

Le document en format XML

<record>
<TEI>
<teiHeader>
<fileDesc>
<titleStmt>
<title xml:lang="en">Online Pairwise Learning Algorithms.</title>
<author>
<name sortKey="Ying, Yiming" sort="Ying, Yiming" uniqKey="Ying Y" first="Yiming" last="Ying">Yiming Ying</name>
<affiliation wicri:level="2">
<nlm:affiliation>Department of Mathematics and Statistics, State University of New York at Albany, Albany, NY 12222, U.S.A. yying@albany.edu.</nlm:affiliation>
<country xml:lang="fr">États-Unis</country>
<wicri:regionArea>Department of Mathematics and Statistics, State University of New York at Albany, Albany, NY 12222</wicri:regionArea>
<placeName>
<region type="state">État de New York</region>
</placeName>
</affiliation>
</author>
<author>
<name sortKey="Zhou, Ding Xuan" sort="Zhou, Ding Xuan" uniqKey="Zhou D" first="Ding-Xuan" last="Zhou">Ding-Xuan Zhou</name>
<affiliation wicri:level="1">
<nlm:affiliation>Department of Mathematics, City University of Hong Kong, Kowloon, Hong Kong, China mazhou@cityu.edu.hk.</nlm:affiliation>
<country wicri:rule="url">Hong Kong</country>
<wicri:regionArea>Department of Mathematics, City University of Hong Kong, Kowloon, Hong Kong</wicri:regionArea>
<wicri:noRegion>Hong Kong</wicri:noRegion>
</affiliation>
</author>
</titleStmt>
<publicationStmt>
<idno type="wicri:source">PubMed</idno>
<date when="2016">2016</date>
<idno type="doi">10.1162/NECO_a_00817</idno>
<idno type="RBID">pubmed:26890352</idno>
<idno type="pmid">26890352</idno>
<idno type="wicri:Area/PubMed/Corpus">000014</idno>
<idno type="wicri:Area/PubMed/Curation">000014</idno>
<idno type="wicri:Area/PubMed/Checkpoint">000019</idno>
<idno type="wicri:Area/Ncbi/Merge">000E90</idno>
<idno type="wicri:Area/Ncbi/Curation">000E90</idno>
<idno type="wicri:Area/Ncbi/Checkpoint">000E90</idno>
<idno type="wicri:Area/Main/Merge">000026</idno>
</publicationStmt>
<sourceDesc>
<biblStruct>
<analytic>
<title xml:lang="en">Online Pairwise Learning Algorithms.</title>
<author>
<name sortKey="Ying, Yiming" sort="Ying, Yiming" uniqKey="Ying Y" first="Yiming" last="Ying">Yiming Ying</name>
<affiliation wicri:level="2">
<nlm:affiliation>Department of Mathematics and Statistics, State University of New York at Albany, Albany, NY 12222, U.S.A. yying@albany.edu.</nlm:affiliation>
<country xml:lang="fr">États-Unis</country>
<wicri:regionArea>Department of Mathematics and Statistics, State University of New York at Albany, Albany, NY 12222</wicri:regionArea>
<placeName>
<region type="state">État de New York</region>
</placeName>
</affiliation>
</author>
<author>
<name sortKey="Zhou, Ding Xuan" sort="Zhou, Ding Xuan" uniqKey="Zhou D" first="Ding-Xuan" last="Zhou">Ding-Xuan Zhou</name>
<affiliation wicri:level="1">
<nlm:affiliation>Department of Mathematics, City University of Hong Kong, Kowloon, Hong Kong, China mazhou@cityu.edu.hk.</nlm:affiliation>
<country wicri:rule="url">Hong Kong</country>
<wicri:regionArea>Department of Mathematics, City University of Hong Kong, Kowloon, Hong Kong</wicri:regionArea>
<wicri:noRegion>Hong Kong</wicri:noRegion>
</affiliation>
</author>
</analytic>
<series>
<title level="j">Neural computation</title>
<idno type="e-ISSN">1530-888X</idno>
<imprint>
<date when="2016" type="published">2016</date>
</imprint>
</series>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc>
<textClass></textClass>
</profileDesc>
</teiHeader>
<front>
<div type="abstract" xml:lang="en">Pairwise learning usually refers to a learning task that involves a loss function depending on pairs of examples, among which the most notable ones are bipartite ranking, metric learning, and AUC maximization. In this letter we study an online algorithm for pairwise learning with a least-square loss function in an unconstrained setting of a reproducing kernel Hilbert space (RKHS) that we refer to as the Online Pairwise lEaRning Algorithm (OPERA). In contrast to existing works (Kar, Sriperumbudur, Jain, & Karnick, 2013 ; Wang, Khardon, Pechyony, & Jones, 2012 ), which require that the iterates are restricted to a bounded domain or the loss function is strongly convex, OPERA is associated with a non-strongly convex objective function and learns the target function in an unconstrained RKHS. Specifically, we establish a general theorem that guarantees the almost sure convergence for the last iterate of OPERA without any assumptions on the underlying distribution. Explicit convergence rates are derived under the condition of polynomially decaying step sizes. We also establish an interesting property for a family of widely used kernels in the setting of pairwise learning and illustrate the convergence results using such kernels. Our methodology mainly depends on the characterization of RKHSs using its associated integral operators and probability inequalities for random variables with values in a Hilbert space.</div>
</front>
</TEI>
</record>

Pour manipuler ce document sous Unix (Dilib)

EXPLOR_STEP=$WICRI_ROOT/Wicri/Musique/explor/OperaV1/Data/Main/Merge
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000026 | SxmlIndent | more

Ou

HfdSelect -h $EXPLOR_AREA/Data/Main/Merge/biblio.hfd -nk 000026 | SxmlIndent | more

Pour mettre un lien sur cette page dans le réseau Wicri

{{Explor lien
   |wiki=    Wicri/Musique
   |area=    OperaV1
   |flux=    Main
   |étape=   Merge
   |type=    RBID
   |clé=     pubmed:26890352
   |texte=   Online Pairwise Learning Algorithms.
}}

Pour générer des pages wiki

HfdIndexSelect -h $EXPLOR_AREA/Data/Main/Merge/RBID.i   -Sk "pubmed:26890352" \
       | HfdSelect -Kh $EXPLOR_AREA/Data/Main/Merge/biblio.hfd   \
       | NlmPubMed2Wicri -a OperaV1 

Wicri

This area was generated with Dilib version V0.6.21.
Data generation: Thu Apr 14 14:59:05 2016. Site generation: Thu Oct 8 06:48:41 2020