The Problem of Overlearning in High-Order ICA Approaches: Analysis and Solutions
Identifieur interne : 000279 ( Main/Exploration ); précédent : 000278; suivant : 000280The Problem of Overlearning in High-Order ICA Approaches: Analysis and Solutions
Auteurs : Jaakko S Rel [Finlande] ; Ricardo Vigário [Allemagne, Finlande]Source :
- Lecture Notes in Computer Science [ 0302-9743 ] ; 2001.
Abstract
Abstract: We consider a type of overlearning typical of independent component analysis algorithms. These can be seen to minimize the mutual information between source estimates. The overlearning causes spikelike signals if there are too few samples or there is a considerable amount of noise present. It is argued that if the data has flicker noise the problem is more severe and is better characterized by bumps instead of spikes. The problem is demonstrated using recorded magnetoencephalographic signals. Several methods are suggested that attempt to solve the overlearning problem or, at least, diminish reduce its effects.
Url:
DOI: 10.1007/3-540-45723-2_99
Affiliations:
Links toward previous steps (curation, corpus...)
Le document en format XML
<record><TEI wicri:istexFullTextTei="biblStruct"><teiHeader><fileDesc><titleStmt><title xml:lang="en">The Problem of Overlearning in High-Order ICA Approaches: Analysis and Solutions</title><author><name sortKey="S Rel, Jaakko" sort="S Rel, Jaakko" uniqKey="S Rel J" first="Jaakko" last="S Rel">Jaakko S Rel</name></author><author><name sortKey="Vigario, Ricardo" sort="Vigario, Ricardo" uniqKey="Vigario R" first="Ricardo" last="Vigário">Ricardo Vigário</name></author></titleStmt><publicationStmt><idno type="wicri:source">ISTEX</idno><idno type="RBID">ISTEX:EED15B332943C34B00EA62FAC243875440F299CB</idno><date when="2001" year="2001">2001</date><idno type="doi">10.1007/3-540-45723-2_99</idno><idno type="url">https://api.istex.fr/document/EED15B332943C34B00EA62FAC243875440F299CB/fulltext/pdf</idno><idno type="wicri:Area/Main/Corpus">000806</idno><idno type="wicri:explorRef" wicri:stream="Main" wicri:step="Corpus" wicri:corpus="ISTEX">000806</idno><idno type="wicri:Area/Main/Curation">000806</idno><idno type="wicri:Area/Main/Exploration">000279</idno><idno type="wicri:explorRef" wicri:stream="Main" wicri:step="Exploration">000279</idno></publicationStmt><sourceDesc><biblStruct><analytic><title level="a" type="main" xml:lang="en">The Problem of Overlearning in High-Order ICA Approaches: Analysis and Solutions</title><author><name sortKey="S Rel, Jaakko" sort="S Rel, Jaakko" uniqKey="S Rel J" first="Jaakko" last="S Rel">Jaakko S Rel</name><affiliation wicri:level="1"><country xml:lang="fr">Finlande</country><wicri:regionArea>Neural Network Research Centre, Helsinki University of Technology, FIN - 02015, HUT</wicri:regionArea><wicri:noRegion>HUT</wicri:noRegion></affiliation><affiliation wicri:level="1"><country wicri:rule="url">Finlande</country></affiliation></author><author><name sortKey="Vigario, Ricardo" sort="Vigario, Ricardo" uniqKey="Vigario R" first="Ricardo" last="Vigário">Ricardo Vigário</name><affiliation wicri:level="3"><country xml:lang="fr">Allemagne</country><wicri:regionArea>GMD - FIRST, D- 12489, Kekuléstrasse 7, Berlin</wicri:regionArea><placeName><region type="land" nuts="3">Berlin</region><settlement type="city">Berlin</settlement></placeName></affiliation><affiliation wicri:level="1"><country wicri:rule="url">Finlande</country></affiliation></author></analytic><monogr></monogr><series><title level="s">Lecture Notes in Computer Science</title><imprint><date>2001</date></imprint><idno type="ISSN">0302-9743</idno><idno type="ISSN">0302-9743</idno></series><idno type="istex">EED15B332943C34B00EA62FAC243875440F299CB</idno><idno type="DOI">10.1007/3-540-45723-2_99</idno><idno type="ChapterID">99</idno><idno type="ChapterID">Chap99</idno></biblStruct></sourceDesc><seriesStmt><idno type="ISSN">0302-9743</idno></seriesStmt></fileDesc><profileDesc><textClass></textClass><langUsage><language ident="en">en</language></langUsage></profileDesc></teiHeader><front><div type="abstract" xml:lang="en">Abstract: We consider a type of overlearning typical of independent component analysis algorithms. These can be seen to minimize the mutual information between source estimates. The overlearning causes spikelike signals if there are too few samples or there is a considerable amount of noise present. It is argued that if the data has flicker noise the problem is more severe and is better characterized by bumps instead of spikes. The problem is demonstrated using recorded magnetoencephalographic signals. Several methods are suggested that attempt to solve the overlearning problem or, at least, diminish reduce its effects.</div></front></TEI><affiliations><list><country><li>Allemagne</li><li>Finlande</li></country><region><li>Berlin</li></region><settlement><li>Berlin</li></settlement></list><tree><country name="Finlande"><noRegion><name sortKey="S Rel, Jaakko" sort="S Rel, Jaakko" uniqKey="S Rel J" first="Jaakko" last="S Rel">Jaakko S Rel</name></noRegion><name sortKey="S Rel, Jaakko" sort="S Rel, Jaakko" uniqKey="S Rel J" first="Jaakko" last="S Rel">Jaakko S Rel</name><name sortKey="Vigario, Ricardo" sort="Vigario, Ricardo" uniqKey="Vigario R" first="Ricardo" last="Vigário">Ricardo Vigário</name></country><country name="Allemagne"><region name="Berlin"><name sortKey="Vigario, Ricardo" sort="Vigario, Ricardo" uniqKey="Vigario R" first="Ricardo" last="Vigário">Ricardo Vigário</name></region></country></tree></affiliations></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/France/explor/AussoisV1/Data/Main/Exploration
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000279 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 000279 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Wicri/France
|area= AussoisV1
|flux= Main
|étape= Exploration
|type= RBID
|clé= ISTEX:EED15B332943C34B00EA62FAC243875440F299CB
|texte= The Problem of Overlearning in High-Order ICA Approaches: Analysis and Solutions
}}
| This area was generated with Dilib version V0.6.29. |  |