A new algorithm for learning in piecewise-linear neural networks
Identifieur interne : 000542 ( Main/Exploration ); précédent : 000541; suivant : 000543A new algorithm for learning in piecewise-linear neural networks
Auteurs : E. F. Gad [Canada] ; A. F. Atiya [Canada, États-Unis] ; S. Shaheen [Canada, Égypte] ; A. El-Dessouki [Égypte]Source :
- Neural Networks [ 0893-6080 ] ; 2000.
English descriptors
- KwdEn :
Abstract
Piecewise-linear (PWL) neural networks are widely known for their amenability to digital implementation. This paper presents a new algorithm for learning in PWL networks consisting of a single hidden layer. The approach adopted is based upon constructing a continuous PWL error function and developing an efficient algorithm to minimize it. The algorithm consists of two basic stages in searching the weight space. The first stage of the optimization algorithm is used to locate a point in the weight space representing the intersection of N linearly independent hyperplanes, with N being the number of weights in the network. The second stage is then called to use this point as a starting point in order to continue searching by moving along the single-dimension boundaries between the different linear regions of the error function, hopping from one point (representing the intersection of N hyperplanes) to another. The proposed algorithm exhibits significantly accelerated convergence, as compared to standard algorithms such as back-propagation and improved versions of it, such as the conjugate gradient algorithm. In addition, it has the distinct advantage that there are no parameters to adjust, and therefore there is no time-consuming parameters tuning step. The new algorithm is expected to find applications in function approximation, time series prediction and binary classification problems.
Url:
DOI: 10.1016/S0893-6080(00)00024-1
Affiliations:
Links toward previous steps (curation, corpus...)
Le document en format XML
<record><TEI wicri:istexFullTextTei="biblStruct"><teiHeader><fileDesc><titleStmt><title xml:lang="en">A new algorithm for learning in piecewise-linear neural networks</title><author><name sortKey="Gad, E F" sort="Gad, E F" uniqKey="Gad E" first="E. F." last="Gad">E. F. Gad</name></author><author><name sortKey="Atiya, A F" sort="Atiya, A F" uniqKey="Atiya A" first="A. F." last="Atiya">A. F. Atiya</name></author><author><name sortKey="Shaheen, S" sort="Shaheen, S" uniqKey="Shaheen S" first="S." last="Shaheen">S. Shaheen</name></author><author><name sortKey="El Dessouki, A" sort="El Dessouki, A" uniqKey="El Dessouki A" first="A." last="El-Dessouki">A. El-Dessouki</name></author></titleStmt><publicationStmt><idno type="wicri:source">ISTEX</idno><idno type="RBID">ISTEX:0A0F6E562C7EBAD116E20CAAFC45769948A76E32</idno><date when="2000" year="2000">2000</date><idno type="doi">10.1016/S0893-6080(00)00024-1</idno><idno type="url">https://api.istex.fr/document/0A0F6E562C7EBAD116E20CAAFC45769948A76E32/fulltext/pdf</idno><idno type="wicri:Area/Main/Corpus">000659</idno><idno type="wicri:Area/Main/Curation">000659</idno><idno type="wicri:Area/Main/Exploration">000542</idno><idno type="wicri:explorRef" wicri:stream="Main" wicri:step="Exploration">000542</idno></publicationStmt><sourceDesc><biblStruct><analytic><title level="a" type="main" xml:lang="en">A new algorithm for learning in piecewise-linear neural networks</title><author><name sortKey="Gad, E F" sort="Gad, E F" uniqKey="Gad E" first="E. F." last="Gad">E. F. Gad</name><affiliation wicri:level="1"><country wicri:rule="url">Canada</country></affiliation><affiliation wicri:level="1"><country>Canada</country><wicri:regionArea>Department of Electrical Engineering, Carleton University, 1125 Colonel By Drive, Ottawa, Ont.</wicri:regionArea><wicri:noRegion>Ont.</wicri:noRegion></affiliation></author><author><name sortKey="Atiya, A F" sort="Atiya, A F" uniqKey="Atiya A" first="A. F." last="Atiya">A. F. Atiya</name><affiliation wicri:level="1"><country wicri:rule="url">Canada</country></affiliation><affiliation wicri:level="1"><country xml:lang="fr">États-Unis</country><wicri:regionArea>Department of Electrical Engineering, Caltech 136-93 Pasadena, CA 91125</wicri:regionArea><wicri:noRegion>CA 91125</wicri:noRegion></affiliation></author><author><name sortKey="Shaheen, S" sort="Shaheen, S" uniqKey="Shaheen S" first="S." last="Shaheen">S. Shaheen</name><affiliation wicri:level="1"><country wicri:rule="url">Canada</country></affiliation><affiliation wicri:level="1"><country xml:lang="fr">Égypte</country><wicri:regionArea>Department of Computer Engineering, Cairo University, Giza</wicri:regionArea><wicri:noRegion>Giza</wicri:noRegion></affiliation></author><author><name sortKey="El Dessouki, A" sort="El Dessouki, A" uniqKey="El Dessouki A" first="A." last="El-Dessouki">A. El-Dessouki</name><affiliation wicri:level="1"><country xml:lang="fr">Égypte</country><wicri:regionArea>Informatics Research Institute, MCSRTA, Alexandria</wicri:regionArea><wicri:noRegion>Alexandria</wicri:noRegion></affiliation></author></analytic><monogr></monogr><series><title level="j">Neural Networks</title><title level="j" type="abbrev">NN</title><idno type="ISSN">0893-6080</idno><imprint><publisher>ELSEVIER</publisher><date type="published" when="2000">2000</date><biblScope unit="volume">13</biblScope><biblScope unit="issue">4–5</biblScope><biblScope unit="page" from="485">485</biblScope><biblScope unit="page" to="505">505</biblScope></imprint><idno type="ISSN">0893-6080</idno></series><idno type="istex">0A0F6E562C7EBAD116E20CAAFC45769948A76E32</idno><idno type="DOI">10.1016/S0893-6080(00)00024-1</idno><idno type="PII">S0893-6080(00)00024-1</idno></biblStruct></sourceDesc><seriesStmt><idno type="ISSN">0893-6080</idno></seriesStmt></fileDesc><profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>Convergence</term><term>Function approximation</term><term>Neural networks</term></keywords></textClass><langUsage><language ident="en">en</language></langUsage></profileDesc></teiHeader><front><div type="abstract" xml:lang="en">Piecewise-linear (PWL) neural networks are widely known for their amenability to digital implementation. This paper presents a new algorithm for learning in PWL networks consisting of a single hidden layer. The approach adopted is based upon constructing a continuous PWL error function and developing an efficient algorithm to minimize it. The algorithm consists of two basic stages in searching the weight space. The first stage of the optimization algorithm is used to locate a point in the weight space representing the intersection of N linearly independent hyperplanes, with N being the number of weights in the network. The second stage is then called to use this point as a starting point in order to continue searching by moving along the single-dimension boundaries between the different linear regions of the error function, hopping from one point (representing the intersection of N hyperplanes) to another. The proposed algorithm exhibits significantly accelerated convergence, as compared to standard algorithms such as back-propagation and improved versions of it, such as the conjugate gradient algorithm. In addition, it has the distinct advantage that there are no parameters to adjust, and therefore there is no time-consuming parameters tuning step. The new algorithm is expected to find applications in function approximation, time series prediction and binary classification problems.</div></front></TEI><affiliations><list><country><li>Canada</li><li>Égypte</li><li>États-Unis</li></country></list><tree><country name="Canada"><noRegion><name sortKey="Gad, E F" sort="Gad, E F" uniqKey="Gad E" first="E. F." last="Gad">E. F. Gad</name></noRegion><name sortKey="Atiya, A F" sort="Atiya, A F" uniqKey="Atiya A" first="A. F." last="Atiya">A. F. Atiya</name><name sortKey="Gad, E F" sort="Gad, E F" uniqKey="Gad E" first="E. F." last="Gad">E. F. Gad</name><name sortKey="Shaheen, S" sort="Shaheen, S" uniqKey="Shaheen S" first="S." last="Shaheen">S. Shaheen</name></country><country name="États-Unis"><noRegion><name sortKey="Atiya, A F" sort="Atiya, A F" uniqKey="Atiya A" first="A. F." last="Atiya">A. F. Atiya</name></noRegion></country><country name="Égypte"><noRegion><name sortKey="Shaheen, S" sort="Shaheen, S" uniqKey="Shaheen S" first="S." last="Shaheen">S. Shaheen</name></noRegion><name sortKey="El Dessouki, A" sort="El Dessouki, A" uniqKey="El Dessouki A" first="A." last="El-Dessouki">A. El-Dessouki</name></country></tree></affiliations></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Amerique/explor/CaltechV1/Data/Main/Exploration
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000542 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 000542 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Wicri/Amerique
|area= CaltechV1
|flux= Main
|étape= Exploration
|type= RBID
|clé= ISTEX:0A0F6E562C7EBAD116E20CAAFC45769948A76E32
|texte= A new algorithm for learning in piecewise-linear neural networks
}}
| This area was generated with Dilib version V0.6.32. |  |