Publicação

Pruning weightless neural networks

Ver documento

Detalhes bibliográficos
Resumo:Weightless neural networks (WNNs) are a type of machine learning model which perform prediction using lookup tables (LUTs) instead of arithmetic operations. Recent advancements in WNNs have reduced model sizes and improved accuracies, reducing the gap in accuracy with deep neural networks (DNNs). Modern DNNs leverage “pruning” techniques to reduce model size, but this has not previously been explored for WNNs. We propose a WNN pruning strategy based on identifying and culling the LUTs which contribute least to overall model accuracy. We demonstrate an average 40% reduction in model size with at most 1% reduction in accuracy.
Autores principais:Susskind, Z.
Outros Autores:Bacellar, A. T. L.; Arora, A.; Villon, L. A. Q.; Mendanha, R.; Araújo, L. S. de.; Dutra, D. L. C.; Lima, P. M. V.; França, F. M. G.; Miranda, I. D. S.; Breternitz Jr., M.; John, L. K.
Ano:2022
País:Portugal
Tipo de documento:documento de conferência
Tipo de acesso:acesso aberto
Instituição associada:ISCTE
Idioma:inglês
Origem:Repositório ISCTE
_version_ 1868443496203943936
author Susskind, Z.
author2 Bacellar, A. T. L.
Arora, A.
Villon, L. A. Q.
Mendanha, R.
Araújo, L. S. de.
Dutra, D. L. C.
Lima, P. M. V.
França, F. M. G.
Miranda, I. D. S.
Breternitz Jr., M.
John, L. K.
author2_role author
author
author
author
author
author
author
author
author
author
author
author_facet Susskind, Z.
Bacellar, A. T. L.
Arora, A.
Villon, L. A. Q.
Mendanha, R.
Araújo, L. S. de.
Dutra, D. L. C.
Lima, P. M. V.
França, F. M. G.
Miranda, I. D. S.
Breternitz Jr., M.
John, L. K.
author_role author
country_str PT
creators_json_txt [{\"Person.name\":\"Susskind, Z.\"},{\"Person.name\":\"Bacellar, A. T. L.\"},{\"Person.name\":\"Arora, A.\"},{\"Person.name\":\"Villon, L. A. Q.\"},{\"Person.name\":\"Mendanha, R.\"},{\"Person.name\":\"Araújo, L. S. de.\"},{\"Person.name\":\"Dutra, D. L. C.\"},{\"Person.name\":\"Lima, P. M. V.\"},{\"Person.name\":\"França, F. M. G.\"},{\"Person.name\":\"Miranda, I. D. S.\"},{\"Person.name\":\"Breternitz Jr., M.\"},{\"Person.name\":\"John, L. K.\"}]
datacite.creators.creator.creatorName.fl_str_mv Susskind, Z.
Bacellar, A. T. L.
Arora, A.
Villon, L. A. Q.
Mendanha, R.
Araújo, L. S. de.
Dutra, D. L. C.
Lima, P. M. V.
França, F. M. G.
Miranda, I. D. S.
Breternitz Jr., M.
John, L. K.
datacite.date.Accepted.fl_str_mv 2022-01-01T00:00:00Z
datacite.date.available.fl_str_mv 2022-12-05T12:31:40Z
datacite.date.embargoed.fl_str_mv 2022-12-05T12:31:40Z
datacite.rights.fl_str_mv http://purl.org/coar/access_right/c_abf2
datacite.titles.title.fl_str_mv Pruning weightless neural networks
dc.creator.none.fl_str_mv Susskind, Z.
Bacellar, A. T. L.
Arora, A.
Villon, L. A. Q.
Mendanha, R.
Araújo, L. S. de.
Dutra, D. L. C.
Lima, P. M. V.
França, F. M. G.
Miranda, I. D. S.
Breternitz Jr., M.
John, L. K.
dc.date.Accepted.fl_str_mv 2022-01-01T00:00:00Z
dc.date.available.fl_str_mv 2022-12-05T12:31:40Z
dc.date.embargoed.fl_str_mv 2022-12-05T12:31:40Z
dc.format.none.fl_str_mv application/pdf
dc.identifier.none.fl_str_mv http://hdl.handle.net/10071/26522
dc.language.none.fl_str_mv eng
dc.publisher.none.fl_str_mv ESANN
dc.rights.none.fl_str_mv http://purl.org/coar/access_right/c_abf2
dc.title.fl_str_mv Pruning weightless neural networks
dc.type.none.fl_str_mv http://purl.org/coar/resource_type/c_c94f
description Weightless neural networks (WNNs) are a type of machine learning model which perform prediction using lookup tables (LUTs) instead of arithmetic operations. Recent advancements in WNNs have reduced model sizes and improved accuracies, reducing the gap in accuracy with deep neural networks (DNNs). Modern DNNs leverage “pruning” techniques to reduce model size, but this has not previously been explored for WNNs. We propose a WNN pruning strategy based on identifying and culling the LUTs which contribute least to overall model accuracy. We demonstrate an average 40% reduction in model size with at most 1% reduction in accuracy.
dirty 0
eu_rights_str_mv openAccess
format conferenceObject
id iscte_38bcea3f3dee55bedd5de024b3d41a4e
identifier.url.fl_str_mv http://hdl.handle.net/10071/26522
instacron_str iscte
institution ISCTE
instname_str ISCTE
language eng
network_acronym_str iscte
network_name_str Repositório ISCTE
oai_identifier_str oai:repositorio.iscte-iul.pt:10071/26522
organization_str_mv urn:organizationAcronym:iscte
person_str_mv Susskind, Z.
Bacellar, A. T. L.
Arora, A.
Villon, L. A. Q.
Mendanha, R.
Araújo, L. S. de.
Dutra, D. L. C.
Lima, P. M. V.
França, F. M. G.
Miranda, I. D. S.
Breternitz Jr., M.
John, L. K.
publishDate 2022
publisher.none.fl_str_mv ESANN
reponame_str Repositório ISCTE
repository_id_str urn:repositoryAcronym:iscte
service_str_mv urn:repositoryAcronym:iscte
spelling engWeightless neural networks (WNNs) are a type of machine learning model which perform prediction using lookup tables (LUTs) instead of arithmetic operations. Recent advancements in WNNs have reduced model sizes and improved accuracies, reducing the gap in accuracy with deep neural networks (DNNs). Modern DNNs leverage “pruning” techniques to reduce model size, but this has not previously been explored for WNNs. We propose a WNN pruning strategy based on identifying and culling the LUTs which contribute least to overall model accuracy. We demonstrate an average 40% reduction in model size with at most 1% reduction in accuracy.application/pdfengESANNengPruning weightless neural networksSusskind, Z.Bacellar, A. T. L.Arora, A.Villon, L. A. Q.Mendanha, R.Araújo, L. S. de.Dutra, D. L. C.Lima, P. M. V.França, F. M. G.Miranda, I. D. S.Breternitz Jr., M.John, L. K.Handlehttp://hdl.handle.net/10071/26522ISBNIsPartOf978287587084-1DOIIsPartOf10.14428/esann/2022.ES2022-552022-12-05T12:31:40Z2022-01-01T00:00:00Z20222022-12-05T12:27:28Zhttp://purl.org/coar/access_right/c_abf2open access1475404 byteshttp://purl.org/coar/access_right/c_abf2application/pdffulltexthttps://repositorio.iscte-iul.pt/bitstreams/c87ac2c2-9bde-48ed-9658-8729b336ca96/downloadother research producthttp://purl.org/coar/resource_type/c_c94fconference object
spellingShingle Pruning weightless neural networks
Susskind, Z.
status SINGLETON
title Pruning weightless neural networks
title_full Pruning weightless neural networks
title_fullStr Pruning weightless neural networks
title_full_unstemmed Pruning weightless neural networks
title_short Pruning weightless neural networks
title_sort Pruning weightless neural networks
url http://hdl.handle.net/10071/26522
visible 1