Beyond Boosting: Recursive ECOC Learning Machines
Identifieur interne : 004814 ( Istex/Corpus ); précédent : 004813; suivant : 004815Beyond Boosting: Recursive ECOC Learning Machines
Auteurs : Elizabeth Tapia ; José C. González ; Alexander Hütermann ; Javier GarcíaSource :
- [ 0302-9743 ]
Abstract
Abstract: We present a wide experimental work evaluating the behaviour of Recursive ECOC (RECOC) [1] learning machines based on Low Density Parity Check (LDPC) coding structures. We show that owing to the iterative decoding algorithms behind LDPC codes, RECOC multiclass learning is progressively achieved. This learning behaviour confirms the existence of new boosting dimension, the one provided by the coding space. We present a method for searching potential good RECOC codes from LDPC ones. Starting from a properly selected LDPC code, we assess the effect of boosting in both weak and strong binary learners. For nearly all domains, we find that boosting a strong learner like a Decision Tree is as effective as boosting a weak one like a Decision Stump. This surprising result substantiates the hypothesis that weakening strong classifiers by boosting has a decorrelation effect, which can be used to improve RECOC learning.
Url:
DOI: 10.1007/978-3-540-25966-4_6
Links to Exploration step
ISTEX:C536CA7A9106C957F25EA26EB2C62A846FC927C0Le document en format XML
<record><TEI wicri:istexFullTextTei="no"><teiHeader><fileDesc><titleStmt><title xml:lang="en">Beyond Boosting: Recursive ECOC Learning Machines</title><author><name sortKey="Tapia, Elizabeth" sort="Tapia, Elizabeth" uniqKey="Tapia E" first="Elizabeth" last="Tapia">Elizabeth Tapia</name><affiliation><mods:affiliation>Department of Electronic Engineering, National University of Rosario, Argentina</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: etapia@eie.fceia.unr.edu.ar</mods:affiliation></affiliation></author><author><name sortKey="Gonzalez, Jose C" sort="Gonzalez, Jose C" uniqKey="Gonzalez J" first="José C." last="González">José C. González</name><affiliation><mods:affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: jgonzalez@gsi.dit.upm.es</mods:affiliation></affiliation></author><author><name sortKey="Hutermann, Alexander" sort="Hutermann, Alexander" uniqKey="Hutermann A" first="Alexander" last="Hütermann">Alexander Hütermann</name><affiliation><mods:affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: ahutermann@gsi.dit.upm.es</mods:affiliation></affiliation></author><author><name sortKey="Garcia, Javier" sort="Garcia, Javier" uniqKey="Garcia J" first="Javier" last="García">Javier García</name><affiliation><mods:affiliation>Department of Informatic Systems, Complutense University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: javiergv@sip.ucm.es</mods:affiliation></affiliation></author></titleStmt><publicationStmt><idno type="wicri:source">ISTEX</idno><idno type="RBID">ISTEX:C536CA7A9106C957F25EA26EB2C62A846FC927C0</idno><date when="2004" year="2004">2004</date><idno type="doi">10.1007/978-3-540-25966-4_6</idno><idno type="url">https://api.istex.fr/document/C536CA7A9106C957F25EA26EB2C62A846FC927C0/fulltext/pdf</idno><idno type="wicri:Area/Istex/Corpus">004814</idno><idno type="wicri:explorRef" wicri:stream="Istex" wicri:step="Corpus" wicri:corpus="ISTEX">004814</idno></publicationStmt><sourceDesc><biblStruct><analytic><author><name sortKey="Tapia, Elizabeth" sort="Tapia, Elizabeth" uniqKey="Tapia E" first="Elizabeth" last="Tapia">Elizabeth Tapia</name><affiliation><mods:affiliation>Department of Electronic Engineering, National University of Rosario, Argentina</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: etapia@eie.fceia.unr.edu.ar</mods:affiliation></affiliation></author><author><name sortKey="Gonzalez, Jose C" sort="Gonzalez, Jose C" uniqKey="Gonzalez J" first="José C." last="González">José C. González</name><affiliation><mods:affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: jgonzalez@gsi.dit.upm.es</mods:affiliation></affiliation></author><author><name sortKey="Hutermann, Alexander" sort="Hutermann, Alexander" uniqKey="Hutermann A" first="Alexander" last="Hütermann">Alexander Hütermann</name><affiliation><mods:affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: ahutermann@gsi.dit.upm.es</mods:affiliation></affiliation></author><author><name sortKey="Garcia, Javier" sort="Garcia, Javier" uniqKey="Garcia J" first="Javier" last="García">Javier García</name><affiliation><mods:affiliation>Department of Informatic Systems, Complutense University of Madrid, Spain</mods:affiliation></affiliation><affiliation><mods:affiliation>E-mail: javiergv@sip.ucm.es</mods:affiliation></affiliation></author></analytic><series><idno type="ISSN">0302-9743</idno></series></biblStruct></sourceDesc><seriesStmt><idno type="ISSN">0302-9743</idno></seriesStmt></fileDesc><profileDesc><textClass></textClass></profileDesc></teiHeader><front><div type="abstract" xml:lang="en">Abstract: We present a wide experimental work evaluating the behaviour of Recursive ECOC (RECOC) [1] learning machines based on Low Density Parity Check (LDPC) coding structures. We show that owing to the iterative decoding algorithms behind LDPC codes, RECOC multiclass learning is progressively achieved. This learning behaviour confirms the existence of new boosting dimension, the one provided by the coding space. We present a method for searching potential good RECOC codes from LDPC ones. Starting from a properly selected LDPC code, we assess the effect of boosting in both weak and strong binary learners. For nearly all domains, we find that boosting a strong learner like a Decision Tree is as effective as boosting a weak one like a Decision Stump. This surprising result substantiates the hypothesis that weakening strong classifiers by boosting has a decorrelation effect, which can be used to improve RECOC learning.</div></front></TEI><istex errorEnrichmentsRefBib="curl or Istex failure"><corpusName>springer</corpusName><author><json:item><name>Elizabeth Tapia</name><affiliations><json:string>Department of Electronic Engineering, National University of Rosario, Argentina</json:string><json:string>E-mail: etapia@eie.fceia.unr.edu.ar</json:string></affiliations></json:item><json:item><name>José C. González</name><affiliations><json:string>Department of Telematic Engineering, Technical University of Madrid, Spain</json:string><json:string>E-mail: jgonzalez@gsi.dit.upm.es</json:string></affiliations></json:item><json:item><name>Alexander Hütermann</name><affiliations><json:string>Department of Telematic Engineering, Technical University of Madrid, Spain</json:string><json:string>E-mail: ahutermann@gsi.dit.upm.es</json:string></affiliations></json:item><json:item><name>Javier García</name><affiliations><json:string>Department of Informatic Systems, Complutense University of Madrid, Spain</json:string><json:string>E-mail: javiergv@sip.ucm.es</json:string></affiliations></json:item></author><language><json:string>eng</json:string></language><abstract>Abstract: We present a wide experimental work evaluating the behaviour of Recursive ECOC (RECOC) [1] learning machines based on Low Density Parity Check (LDPC) coding structures. We show that owing to the iterative decoding algorithms behind LDPC codes, RECOC multiclass learning is progressively achieved. This learning behaviour confirms the existence of new boosting dimension, the one provided by the coding space. We present a method for searching potential good RECOC codes from LDPC ones. Starting from a properly selected LDPC code, we assess the effect of boosting in both weak and strong binary learners. For nearly all domains, we find that boosting a strong learner like a Decision Tree is as effective as boosting a weak one like a Decision Stump. This surprising result substantiates the hypothesis that weakening strong classifiers by boosting has a decorrelation effect, which can be used to improve RECOC learning.</abstract><qualityIndicators><score>5.861</score><pdfVersion>1.3</pdfVersion><pdfPageSize>430 x 660 pts</pdfPageSize><refBibsNative>false</refBibsNative><keywordCount>0</keywordCount><abstractCharCount>931</abstractCharCount><pdfWordCount>4121</pdfWordCount><pdfCharCount>20472</pdfCharCount><pdfPageCount>10</pdfPageCount><abstractWordCount>145</abstractWordCount></qualityIndicators><title>Beyond Boosting: Recursive ECOC Learning Machines</title><genre.original><json:string>OriginalPaper</json:string></genre.original><chapterId><json:string>6</json:string><json:string>Chap6</json:string></chapterId><genre><json:string>conference [eBooks]</json:string></genre><serie><editor><json:item><name>Takeo Kanade</name><affiliations><json:string>Carnegie Mellon University, Pittsburgh, PA, USA</json:string></affiliations></json:item><json:item><name>Josef Kittler</name><affiliations><json:string>University of Surrey, Guildford, UK</json:string></affiliations></json:item><json:item><name>Jon M. Kleinberg</name><affiliations><json:string>Cornell University, Ithaca, NY, USA</json:string></affiliations></json:item><json:item><name>Friedemann Mattern</name><affiliations><json:string>ETH Zurich, Switzerland</json:string></affiliations></json:item><json:item><name>John C. Mitchell</name><affiliations><json:string>Stanford University, CA, USA</json:string></affiliations></json:item><json:item><name>Moni Naor</name><affiliations><json:string>Weizmann Institute of Science, Rehovot, Israel</json:string></affiliations></json:item><json:item><name>Oscar Nierstrasz</name><affiliations><json:string>University of Bern, Switzerland</json:string></affiliations></json:item><json:item><name>C. Pandu Rangan</name><affiliations><json:string>Indian Institute of Technology, Madras, India</json:string></affiliations></json:item><json:item><name>Bernhard Steffen</name><affiliations><json:string>University of Dortmund, Germany</json:string></affiliations></json:item><json:item><name>Madhu Sudan</name><affiliations><json:string>Massachusetts Institute of Technology, MA, USA</json:string></affiliations></json:item><json:item><name>Demetri Terzopoulos</name><affiliations><json:string>New York University, NY, USA</json:string></affiliations></json:item><json:item><name>Dough Tygar</name><affiliations><json:string>University of California, Berkeley, CA, USA</json:string></affiliations></json:item><json:item><name>Moshe Y. Vardi</name><affiliations><json:string>Rice University, Houston, TX, USA</json:string></affiliations></json:item><json:item><name>Gerhard Weikum</name><affiliations><json:string>Max-Planck Institute of Computer Science, Saarbruecken, Germany</json:string></affiliations></json:item></editor><issn><json:string>0302-9743</json:string></issn><language><json:string>unknown</json:string></language><eissn><json:string>1611-3349</json:string></eissn><title>Lecture Notes in Computer Science</title><copyrightDate>2004</copyrightDate></serie><host><editor><json:item><name>Fabio Roli</name><affiliations><json:string>Department of Electrical and Electronic Engineering, Piazza d’Armi, University of Cagliari, 09123, Cagliari, Italy</json:string><json:string>E-mail: roli@diee.unica.it</json:string></affiliations></json:item><json:item><name>Josef Kittler</name><affiliations><json:string>Centre for Vision, Speech and Signal Processing, University of Surrey, GU2 7XH, Guildford, UK</json:string><json:string>E-mail: J.Kittler@surrey.ac.uk</json:string></affiliations></json:item><json:item><name>Terry Windeatt</name><affiliations><json:string>Centre for Vision, Speech and Signal Proc (CVSSP), University of Surrey, GU2 7XH, Guildford, Surrey, United Kingdom</json:string><json:string>E-mail: t.windeatt@surrey.ac.uk</json:string></affiliations></json:item></editor><subject><json:item><value>Computer Science</value></json:item><json:item><value>Computer Science</value></json:item><json:item><value>Pattern Recognition</value></json:item><json:item><value>Image Processing and Computer Vision</value></json:item><json:item><value>Artificial Intelligence (incl. Robotics)</value></json:item><json:item><value>Computation by Abstract Devices</value></json:item></subject><isbn><json:string>978-3-540-22144-9</json:string></isbn><language><json:string>unknown</json:string></language><eissn><json:string>1611-3349</json:string></eissn><title>Multiple Classifier Systems</title><genre.original><json:string>Proceedings</json:string></genre.original><bookId><json:string>978-3-540-25966-4</json:string></bookId><volume>3077</volume><pages><last>71</last><first>62</first></pages><issn><json:string>0302-9743</json:string></issn><genre><json:string>Book Series</json:string></genre><eisbn><json:string>978-3-540-25966-4</json:string></eisbn><copyrightDate>2004</copyrightDate><doi><json:string>10.1007/b98227</json:string></doi></host><publicationDate>2004</publicationDate><copyrightDate>2004</copyrightDate><doi><json:string>10.1007/978-3-540-25966-4_6</json:string></doi><id>C536CA7A9106C957F25EA26EB2C62A846FC927C0</id><score>1</score><fulltext><json:item><original>true</original><mimetype>application/pdf</mimetype><extension>pdf</extension><uri>https://api.istex.fr/document/C536CA7A9106C957F25EA26EB2C62A846FC927C0/fulltext/pdf</uri></json:item><json:item><original>false</original><mimetype>application/zip</mimetype><extension>zip</extension><uri>https://api.istex.fr/document/C536CA7A9106C957F25EA26EB2C62A846FC927C0/fulltext/zip</uri></json:item><json:item><original>false</original><mimetype>text/plain</mimetype><extension>txt</extension><uri>https://api.istex.fr/document/C536CA7A9106C957F25EA26EB2C62A846FC927C0/fulltext/txt</uri></json:item></fulltext><metadata><istex:metadataXml wicri:clean="Springer, Publisher found" wicri:toSee="no header"><istex:xmlDeclaration>version="1.0" encoding="UTF-8"</istex:xmlDeclaration><istex:docType PUBLIC="-//Springer-Verlag//DTD A++ V2.4//EN" URI="http://devel.springer.de/A++/V2.4/DTD/A++V2.4.dtd" name="istex:docType"></istex:docType><istex:document><Publisher><PublisherInfo><PublisherName>Springer Berlin Heidelberg</PublisherName><PublisherLocation>Berlin, Heidelberg</PublisherLocation></PublisherInfo><Series><SeriesInfo SeriesType="Series" TocLevels="0"><SeriesID>558</SeriesID><SeriesPrintISSN>0302-9743</SeriesPrintISSN><SeriesElectronicISSN>1611-3349</SeriesElectronicISSN><SeriesTitle Language="En">Lecture Notes in Computer Science</SeriesTitle></SeriesInfo><SeriesHeader><EditorGroup><Editor AffiliationIDS="Aff1"><EditorName DisplayOrder="Western"><GivenName>Takeo</GivenName><FamilyName>Kanade</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff2"><EditorName DisplayOrder="Western"><GivenName>Josef</GivenName><FamilyName>Kittler</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff3"><EditorName DisplayOrder="Western"><GivenName>Jon</GivenName><GivenName>M.</GivenName><FamilyName>Kleinberg</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff4"><EditorName DisplayOrder="Western"><GivenName>Friedemann</GivenName><FamilyName>Mattern</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff5"><EditorName DisplayOrder="Western"><GivenName>John</GivenName><GivenName>C.</GivenName><FamilyName>Mitchell</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff6"><EditorName DisplayOrder="Western"><GivenName>Moni</GivenName><FamilyName>Naor</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff7"><EditorName DisplayOrder="Western"><GivenName>Oscar</GivenName><FamilyName>Nierstrasz</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff8"><EditorName DisplayOrder="Western"><GivenName>C.</GivenName><FamilyName>Pandu Rangan</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff9"><EditorName DisplayOrder="Western"><GivenName>Bernhard</GivenName><FamilyName>Steffen</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff10"><EditorName DisplayOrder="Western"><GivenName>Madhu</GivenName><FamilyName>Sudan</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff11"><EditorName DisplayOrder="Western"><GivenName>Demetri</GivenName><FamilyName>Terzopoulos</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff12"><EditorName DisplayOrder="Western"><GivenName>Dough</GivenName><FamilyName>Tygar</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff13"><EditorName DisplayOrder="Western"><GivenName>Moshe</GivenName><GivenName>Y.</GivenName><FamilyName>Vardi</FamilyName></EditorName></Editor><Editor AffiliationIDS="Aff14"><EditorName DisplayOrder="Western"><GivenName>Gerhard</GivenName><FamilyName>Weikum</FamilyName></EditorName></Editor><Affiliation ID="Aff1"><OrgName>Carnegie Mellon University</OrgName><OrgAddress><City>Pittsburgh</City><State>PA</State><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff2"><OrgName>University of Surrey</OrgName><OrgAddress><City>Guildford</City><Country>UK</Country></OrgAddress></Affiliation><Affiliation ID="Aff3"><OrgName>Cornell University</OrgName><OrgAddress><City>Ithaca</City><State>NY</State><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff4"><OrgName>ETH Zurich</OrgName><OrgAddress><Country>Switzerland</Country></OrgAddress></Affiliation><Affiliation ID="Aff5"><OrgName>Stanford University</OrgName><OrgAddress><City>CA</City><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff6"><OrgName>Weizmann Institute of Science</OrgName><OrgAddress><City>Rehovot</City><Country>Israel</Country></OrgAddress></Affiliation><Affiliation ID="Aff7"><OrgName>University of Bern</OrgName><OrgAddress><Country>Switzerland</Country></OrgAddress></Affiliation><Affiliation ID="Aff8"><OrgName>Indian Institute of Technology</OrgName><OrgAddress><City>Madras</City><Country>India</Country></OrgAddress></Affiliation><Affiliation ID="Aff9"><OrgName>University of Dortmund</OrgName><OrgAddress><Country>Germany</Country></OrgAddress></Affiliation><Affiliation ID="Aff10"><OrgName>Massachusetts Institute of Technology</OrgName><OrgAddress><City>MA</City><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff11"><OrgName>New York University</OrgName><OrgAddress><City>NY</City><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff12"><OrgName>University of California</OrgName><OrgAddress><City>Berkeley</City><State>CA</State><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff13"><OrgName>Rice University</OrgName><OrgAddress><City>Houston</City><State>TX</State><Country>USA</Country></OrgAddress></Affiliation><Affiliation ID="Aff14"><OrgName>Max-Planck Institute of Computer Science</OrgName><OrgAddress><City>Saarbruecken</City><Country>Germany</Country></OrgAddress></Affiliation></EditorGroup></SeriesHeader><Book Language="En"><BookInfo BookProductType="Proceedings" ContainsESM="No" Language="En" MediaType="eBook" NumberingDepth="2" NumberingStyle="ContentOnly" OutputMedium="All" TocLevels="0"><BookID>978-3-540-25966-4</BookID><BookTitle>Multiple Classifier Systems</BookTitle><BookSubTitle>5th International Workshop, MCS 2004, Cagliari, Italy, June 9-11, 2004. Proceedings</BookSubTitle><BookVolumeNumber>3077</BookVolumeNumber><BookSequenceNumber>3077</BookSequenceNumber><BookDOI>10.1007/b98227</BookDOI><BookTitleID>82942</BookTitleID><BookPrintISBN>978-3-540-22144-9</BookPrintISBN><BookElectronicISBN>978-3-540-25966-4</BookElectronicISBN><BookChapterCount>37</BookChapterCount><BookCopyright><CopyrightHolderName>Springer-Verlag Berlin Heidelberg</CopyrightHolderName><CopyrightYear>2004</CopyrightYear></BookCopyright><BookSubjectGroup><BookSubject Code="I" Type="Primary">Computer Science</BookSubject><BookSubject Code="I2203X" Priority="1" Type="Secondary">Pattern Recognition</BookSubject><BookSubject Code="I22021" Priority="2" Type="Secondary">Image Processing and Computer Vision</BookSubject><BookSubject Code="I21017" Priority="3" Type="Secondary">Artificial Intelligence (incl. Robotics)</BookSubject><BookSubject Code="I16013" Priority="4" Type="Secondary">Computation by Abstract Devices</BookSubject><SubjectCollection Code="SUCO11645">Computer Science</SubjectCollection></BookSubjectGroup><BookContext><SeriesID>558</SeriesID></BookContext></BookInfo><BookHeader><EditorGroup><Editor AffiliationIDS="Aff15"><EditorName DisplayOrder="Western"><GivenName>Fabio</GivenName><FamilyName>Roli</FamilyName></EditorName><Contact><Email>roli@diee.unica.it</Email></Contact></Editor><Editor AffiliationIDS="Aff16"><EditorName DisplayOrder="Western"><GivenName>Josef</GivenName><FamilyName>Kittler</FamilyName></EditorName><Contact><Email>J.Kittler@surrey.ac.uk</Email></Contact></Editor><Editor AffiliationIDS="Aff17"><EditorName DisplayOrder="Western"><GivenName>Terry</GivenName><FamilyName>Windeatt</FamilyName></EditorName><Contact><Email>t.windeatt@surrey.ac.uk</Email></Contact></Editor><Affiliation ID="Aff15"><OrgDivision>Department of Electrical and Electronic Engineering, Piazza d’Armi</OrgDivision><OrgName>University of Cagliari</OrgName><OrgAddress><Postcode>09123</Postcode><City>Cagliari</City><Country>Italy</Country></OrgAddress></Affiliation><Affiliation ID="Aff16"><OrgDivision>Centre for Vision, Speech and Signal Processing</OrgDivision><OrgName>University of Surrey</OrgName><OrgAddress><Postcode>GU2 7XH</Postcode><City>Guildford</City><Country>UK</Country></OrgAddress></Affiliation><Affiliation ID="Aff17"><OrgDivision>Centre for Vision, Speech and Signal Proc (CVSSP)</OrgDivision><OrgName>University of Surrey</OrgName><OrgAddress><Postcode>GU2 7XH</Postcode><City>Guildford</City><State>Surrey</State><Country>United Kingdom</Country></OrgAddress></Affiliation></EditorGroup></BookHeader><Part ID="Part2"><PartInfo TocLevels="0"><PartID>2</PartID><PartSequenceNumber>2</PartSequenceNumber><PartTitle>Bagging and Boosting</PartTitle><PartChapterCount>5</PartChapterCount><PartContext><SeriesID>558</SeriesID><BookTitle>Multiple Classifier Systems</BookTitle></PartContext></PartInfo><Chapter ID="Chap6" Language="En"><ChapterInfo ChapterType="OriginalPaper" ContainsESM="No" NumberingDepth="2" NumberingStyle="ContentOnly" TocLevels="0"><ChapterID>6</ChapterID><ChapterDOI>10.1007/978-3-540-25966-4_6</ChapterDOI><ChapterSequenceNumber>6</ChapterSequenceNumber><ChapterTitle Language="En">Beyond Boosting: Recursive ECOC Learning Machines</ChapterTitle><ChapterFirstPage>62</ChapterFirstPage><ChapterLastPage>71</ChapterLastPage><ChapterCopyright><CopyrightHolderName>Springer-Verlag Berlin Heidelberg</CopyrightHolderName><CopyrightYear>2004</CopyrightYear></ChapterCopyright><ChapterGrants Type="Regular"><MetadataGrant Grant="OpenAccess"></MetadataGrant><AbstractGrant Grant="OpenAccess"></AbstractGrant><BodyPDFGrant Grant="Restricted"></BodyPDFGrant><BodyHTMLGrant Grant="Restricted"></BodyHTMLGrant><BibliographyGrant Grant="Restricted"></BibliographyGrant><ESMGrant Grant="Restricted"></ESMGrant></ChapterGrants><ChapterContext><SeriesID>558</SeriesID><PartID>2</PartID><BookID>978-3-540-25966-4</BookID><BookTitle>Multiple Classifier Systems</BookTitle></ChapterContext></ChapterInfo><ChapterHeader><AuthorGroup><Author AffiliationIDS="Aff18"><AuthorName DisplayOrder="Western"><GivenName>Elizabeth</GivenName><FamilyName>Tapia</FamilyName></AuthorName><Contact><Email>etapia@eie.fceia.unr.edu.ar</Email></Contact></Author><Author AffiliationIDS="Aff19"><AuthorName DisplayOrder="Western"><GivenName>José</GivenName><GivenName>C.</GivenName><FamilyName>González</FamilyName></AuthorName><Contact><Email>jgonzalez@gsi.dit.upm.es</Email></Contact></Author><Author AffiliationIDS="Aff19"><AuthorName DisplayOrder="Western"><GivenName>Alexander</GivenName><FamilyName>Hütermann</FamilyName></AuthorName><Contact><Email>ahutermann@gsi.dit.upm.es</Email></Contact></Author><Author AffiliationIDS="Aff20"><AuthorName DisplayOrder="Western"><GivenName>Javier</GivenName><FamilyName>García</FamilyName></AuthorName><Contact><Email>javiergv@sip.ucm.es</Email></Contact></Author><Affiliation ID="Aff18"><OrgDivision>Department of Electronic Engineering</OrgDivision><OrgName>National University of Rosario</OrgName><OrgAddress><Country>Argentina</Country></OrgAddress></Affiliation><Affiliation ID="Aff19"><OrgDivision>Department of Telematic Engineering</OrgDivision><OrgName>Technical University of Madrid</OrgName><OrgAddress><Country>Spain</Country></OrgAddress></Affiliation><Affiliation ID="Aff20"><OrgDivision>Department of Informatic Systems</OrgDivision><OrgName>Complutense University of Madrid</OrgName><OrgAddress><Country>Spain</Country></OrgAddress></Affiliation></AuthorGroup><Abstract ID="Abs1" Language="En"><Heading>Abstract</Heading><Para>We present a wide experimental work evaluating the behaviour of Recursive ECOC (RECOC) [1] learning machines based on Low Density Parity Check (LDPC) coding structures. We show that owing to the iterative decoding algorithms behind LDPC codes, RECOC multiclass learning is progressively achieved. This learning behaviour confirms the existence of new boosting dimension, the one provided by the coding space. We present a method for searching potential good RECOC codes from LDPC ones. Starting from a properly selected LDPC code, we assess the effect of boosting in both weak and strong binary learners. For nearly all domains, we find that boosting a strong learner like a Decision Tree is as effective as boosting a weak one like a Decision Stump. This surprising result substantiates the hypothesis that weakening strong classifiers by boosting has a decorrelation effect, which can be used to improve RECOC learning.</Para></Abstract></ChapterHeader><NoBody></NoBody></Chapter></Part></Book></Series></Publisher></istex:document></istex:metadataXml><mods version="3.6"><titleInfo lang="en"><title>Beyond Boosting: Recursive ECOC Learning Machines</title></titleInfo><titleInfo type="alternative" contentType="CDATA" lang="en"><title>Beyond Boosting: Recursive ECOC Learning Machines</title></titleInfo><name type="personal"><namePart type="given">Elizabeth</namePart><namePart type="family">Tapia</namePart><affiliation>Department of Electronic Engineering, National University of Rosario, Argentina</affiliation><affiliation>E-mail: etapia@eie.fceia.unr.edu.ar</affiliation><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">José</namePart><namePart type="given">C.</namePart><namePart type="family">González</namePart><affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</affiliation><affiliation>E-mail: jgonzalez@gsi.dit.upm.es</affiliation><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">Alexander</namePart><namePart type="family">Hütermann</namePart><affiliation>Department of Telematic Engineering, Technical University of Madrid, Spain</affiliation><affiliation>E-mail: ahutermann@gsi.dit.upm.es</affiliation><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">Javier</namePart><namePart type="family">García</namePart><affiliation>Department of Informatic Systems, Complutense University of Madrid, Spain</affiliation><affiliation>E-mail: javiergv@sip.ucm.es</affiliation><role><roleTerm type="text">author</roleTerm></role></name><typeOfResource>text</typeOfResource><genre type="conference [eBooks]" displayLabel="OriginalPaper"></genre><originInfo><publisher>Springer Berlin Heidelberg</publisher><place><placeTerm type="text">Berlin, Heidelberg</placeTerm></place><dateIssued encoding="w3cdtf">2004</dateIssued><copyrightDate encoding="w3cdtf">2004</copyrightDate></originInfo><language><languageTerm type="code" authority="rfc3066">en</languageTerm><languageTerm type="code" authority="iso639-2b">eng</languageTerm></language><physicalDescription><internetMediaType>text/html</internetMediaType></physicalDescription><abstract lang="en">Abstract: We present a wide experimental work evaluating the behaviour of Recursive ECOC (RECOC) [1] learning machines based on Low Density Parity Check (LDPC) coding structures. We show that owing to the iterative decoding algorithms behind LDPC codes, RECOC multiclass learning is progressively achieved. This learning behaviour confirms the existence of new boosting dimension, the one provided by the coding space. We present a method for searching potential good RECOC codes from LDPC ones. Starting from a properly selected LDPC code, we assess the effect of boosting in both weak and strong binary learners. For nearly all domains, we find that boosting a strong learner like a Decision Tree is as effective as boosting a weak one like a Decision Stump. This surprising result substantiates the hypothesis that weakening strong classifiers by boosting has a decorrelation effect, which can be used to improve RECOC learning.</abstract><relatedItem type="host"><titleInfo><title>Multiple Classifier Systems</title><subTitle>5th International Workshop, MCS 2004, Cagliari, Italy, June 9-11, 2004. Proceedings</subTitle></titleInfo><name type="personal"><namePart type="given">Fabio</namePart><namePart type="family">Roli</namePart><affiliation>Department of Electrical and Electronic Engineering, Piazza d’Armi, University of Cagliari, 09123, Cagliari, Italy</affiliation><affiliation>E-mail: roli@diee.unica.it</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Josef</namePart><namePart type="family">Kittler</namePart><affiliation>Centre for Vision, Speech and Signal Processing, University of Surrey, GU2 7XH, Guildford, UK</affiliation><affiliation>E-mail: J.Kittler@surrey.ac.uk</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Terry</namePart><namePart type="family">Windeatt</namePart><affiliation>Centre for Vision, Speech and Signal Proc (CVSSP), University of Surrey, GU2 7XH, Guildford, Surrey, United Kingdom</affiliation><affiliation>E-mail: t.windeatt@surrey.ac.uk</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><genre type="Book Series" displayLabel="Proceedings"></genre><originInfo><copyrightDate encoding="w3cdtf">2004</copyrightDate><issuance>monographic</issuance></originInfo><subject><genre>Book Subject Collection</genre><topic authority="SpringerSubjectCodes" authorityURI="SUCO11645">Computer Science</topic></subject><subject><genre>Book Subject Group</genre><topic authority="SpringerSubjectCodes" authorityURI="I">Computer Science</topic><topic authority="SpringerSubjectCodes" authorityURI="I2203X">Pattern Recognition</topic><topic authority="SpringerSubjectCodes" authorityURI="I22021">Image Processing and Computer Vision</topic><topic authority="SpringerSubjectCodes" authorityURI="I21017">Artificial Intelligence (incl. Robotics)</topic><topic authority="SpringerSubjectCodes" authorityURI="I16013">Computation by Abstract Devices</topic></subject><identifier type="DOI">10.1007/b98227</identifier><identifier type="ISBN">978-3-540-22144-9</identifier><identifier type="eISBN">978-3-540-25966-4</identifier><identifier type="ISSN">0302-9743</identifier><identifier type="eISSN">1611-3349</identifier><identifier type="BookTitleID">82942</identifier><identifier type="BookID">978-3-540-25966-4</identifier><identifier type="BookChapterCount">37</identifier><identifier type="BookVolumeNumber">3077</identifier><identifier type="BookSequenceNumber">3077</identifier><identifier type="PartChapterCount">5</identifier><part><date>2004</date><detail type="part"><title>Bagging and Boosting</title></detail><detail type="volume"><number>3077</number><caption>vol.</caption></detail><extent unit="pages"><start>62</start><end>71</end></extent></part><recordInfo><recordOrigin>Springer-Verlag Berlin Heidelberg, 2004</recordOrigin></recordInfo></relatedItem><relatedItem type="series"><titleInfo><title>Lecture Notes in Computer Science</title></titleInfo><name type="personal"><namePart type="given">Takeo</namePart><namePart type="family">Kanade</namePart><affiliation>Carnegie Mellon University, Pittsburgh, PA, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Josef</namePart><namePart type="family">Kittler</namePart><affiliation>University of Surrey, Guildford, UK</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Jon</namePart><namePart type="given">M.</namePart><namePart type="family">Kleinberg</namePart><affiliation>Cornell University, Ithaca, NY, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Friedemann</namePart><namePart type="family">Mattern</namePart><affiliation>ETH Zurich, Switzerland</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">John</namePart><namePart type="given">C.</namePart><namePart type="family">Mitchell</namePart><affiliation>Stanford University, CA, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Moni</namePart><namePart type="family">Naor</namePart><affiliation>Weizmann Institute of Science, Rehovot, Israel</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Oscar</namePart><namePart type="family">Nierstrasz</namePart><affiliation>University of Bern, Switzerland</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">C.</namePart><namePart type="family">Pandu Rangan</namePart><affiliation>Indian Institute of Technology, Madras, India</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Bernhard</namePart><namePart type="family">Steffen</namePart><affiliation>University of Dortmund, Germany</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Madhu</namePart><namePart type="family">Sudan</namePart><affiliation>Massachusetts Institute of Technology, MA, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Demetri</namePart><namePart type="family">Terzopoulos</namePart><affiliation>New York University, NY, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Dough</namePart><namePart type="family">Tygar</namePart><affiliation>University of California, Berkeley, CA, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Moshe</namePart><namePart type="given">Y.</namePart><namePart type="family">Vardi</namePart><affiliation>Rice University, Houston, TX, USA</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><name type="personal"><namePart type="given">Gerhard</namePart><namePart type="family">Weikum</namePart><affiliation>Max-Planck Institute of Computer Science, Saarbruecken, Germany</affiliation><role><roleTerm type="text">editor</roleTerm></role></name><originInfo><copyrightDate encoding="w3cdtf">2004</copyrightDate><issuance>serial</issuance></originInfo><identifier type="ISSN">0302-9743</identifier><identifier type="eISSN">1611-3349</identifier><identifier type="SeriesID">558</identifier><recordInfo><recordOrigin>Springer-Verlag Berlin Heidelberg, 2004</recordOrigin></recordInfo></relatedItem><identifier type="istex">C536CA7A9106C957F25EA26EB2C62A846FC927C0</identifier><identifier type="DOI">10.1007/978-3-540-25966-4_6</identifier><identifier type="ChapterID">6</identifier><identifier type="ChapterID">Chap6</identifier><accessCondition type="use and reproduction" contentType="copyright">Springer-Verlag Berlin Heidelberg, 2004</accessCondition><recordInfo><recordContentSource>SPRINGER</recordContentSource><recordOrigin>Springer-Verlag Berlin Heidelberg, 2004</recordOrigin></recordInfo></mods></metadata></istex></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Ticri/CIDE/explor/TelematiV1/Data/Istex/Corpus
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 004814 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Istex/Corpus/biblio.hfd -nk 004814 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Ticri/CIDE
|area= TelematiV1
|flux= Istex
|étape= Corpus
|type= RBID
|clé= ISTEX:C536CA7A9106C957F25EA26EB2C62A846FC927C0
|texte= Beyond Boosting: Recursive ECOC Learning Machines
}}
| This area was generated with Dilib version V0.6.31. |  |