An investigation into the effects of label noise on dynamic selection algorithms
Autor(a) principal: | |
---|---|
Data de Publicação: | 2020 |
Tipo de documento: | Dissertação |
Idioma: | por |
Título da fonte: | Repositório Institucional da UFPE |
Texto Completo: | https://repositorio.ufpe.br/handle/123456789/37647 |
Resumo: | In the literature on classification problems, it is widely discussed how the presence of label noise can bring about severe degradation in performance. Several works have applied Prototype Selection techniques, Ensemble Methods, or both, in an attempt to alleviate this issue. Nevertheless, these methods are not always able to sufficiently counteract the effects of noise. In this work, we investigate the effects of noise on a particular class of Ensemble Methods, that of Dynamic Selection algorithms, and we are especially interested in the behavior of the Fire-DES++ algorithm, a state of the art algorithm which applies the ENN to algorithm to deal with the effects of noise and imbalance. We propose a method which employs multiple Dynamic Selection sets, based on the Bagging-IH algorithm, which we dub Multiple-Set Dynamic Selection (MSDS), in an attempt to supplant the ENN algorithm on the filtering step. We find that almost all methods based on Dynamic Selection are severely affected by the presence of label noise, with the exception of the KNORAU algorithm. We also find that our proposed method can alleviate the issues caused by noise in some specific scenarios. |
id |
UFPE_f82cd93d4efd88b510bfb775d0b781b2 |
---|---|
oai_identifier_str |
oai:repositorio.ufpe.br:123456789/37647 |
network_acronym_str |
UFPE |
network_name_str |
Repositório Institucional da UFPE |
repository_id_str |
2221 |
spelling |
WALMSLEY, Felipe Nuneshttp://lattes.cnpq.br/8652242028413094http://lattes.cnpq.br/8577312109146354http://lattes.cnpq.br/6269525393139517CAVALCANTI, George Darmiton da CunhaSABOURIN, Robert2020-08-14T17:04:59Z2020-08-14T17:04:59Z2020-01-22WALMSLEY, Felipe Nunes. An investigation into the effects of label noise on dynamic selection algorithms. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020.https://repositorio.ufpe.br/handle/123456789/37647In the literature on classification problems, it is widely discussed how the presence of label noise can bring about severe degradation in performance. Several works have applied Prototype Selection techniques, Ensemble Methods, or both, in an attempt to alleviate this issue. Nevertheless, these methods are not always able to sufficiently counteract the effects of noise. In this work, we investigate the effects of noise on a particular class of Ensemble Methods, that of Dynamic Selection algorithms, and we are especially interested in the behavior of the Fire-DES++ algorithm, a state of the art algorithm which applies the ENN to algorithm to deal with the effects of noise and imbalance. We propose a method which employs multiple Dynamic Selection sets, based on the Bagging-IH algorithm, which we dub Multiple-Set Dynamic Selection (MSDS), in an attempt to supplant the ENN algorithm on the filtering step. We find that almost all methods based on Dynamic Selection are severely affected by the presence of label noise, with the exception of the KNORAU algorithm. We also find that our proposed method can alleviate the issues caused by noise in some specific scenarios.CAPESNa literatura de problemas de classificação, é amplamente discutido como a presença de ruído nos rótulos de classe pode acarretar grave degradação na performance. Vários trabalhos aplicam técnicas de Seleção de Protótipos, Métodos de Ensemble, ou ambos, em uma tentativa de aliviar esse problema. Não obstante, esses métodos nem sempre são capazes de contrabalançar os efeitos do ruído. Neste trabalho, nós investigamos o efeito do ruído em uma classe em particular de Métodos de Ensemble, a classe dos métodos de Seleção Dinâmica, e estamos particularmente interessados no comportamento do algoritmo Fire-DES++, um algoritmo estado da arte que aplica o método Edited Nearest Neighbors (ENN) para lidar com os efeitos de ruído e desbalanceamento. Nós propomos um método que emprega múltiplos conjuntos de Seleção Dinâmica, baseado no algoritmo Bagging-IH, que nós nomeamos Multiple-Set Dynamic Selection (MSDS), em uma tentativa de suplantar o algoritmo ENN no passo de filtragem. Nós observamos que quase todos os métodos baseados em Seleção Dinâmica são fortemente afetados pela presença de ruído, exceto o algoritmo KNORAU. Nós também observamos que, em alguns cenários específicos, o nosso método proposto pode amenizar os problemas causados pelo ruído.porUniversidade Federal de PernambucoPrograma de Pos Graduacao em Ciencia da ComputacaoUFPEBrasilAttribution-NonCommercial-NoDerivs 3.0 Brazilhttp://creativecommons.org/licenses/by-nc-nd/3.0/br/info:eu-repo/semantics/embargoedAccessMétodos de EnsembleSistemas de múltiplos classificadoresSeleção dinâmicaRuído de classeAn investigation into the effects of label noise on dynamic selection algorithmsinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesismestradoreponame:Repositório Institucional da UFPEinstname:Universidade Federal de Pernambuco (UFPE)instacron:UFPEORIGINALDISSERTAÇÃO Felipe Nunes Walmsley.pdfDISSERTAÇÃO Felipe Nunes Walmsley.pdfapplication/pdf722794https://repositorio.ufpe.br/bitstream/123456789/37647/1/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf0b7d873c3459043b8269bc0c0ea1e2d0MD51LICENSElicense.txtlicense.txttext/plain; charset=utf-82310https://repositorio.ufpe.br/bitstream/123456789/37647/3/license.txtbd573a5ca8288eb7272482765f819534MD53TEXTDISSERTAÇÃO Felipe Nunes Walmsley.pdf.txtDISSERTAÇÃO Felipe Nunes Walmsley.pdf.txtExtracted texttext/plain173470https://repositorio.ufpe.br/bitstream/123456789/37647/4/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf.txt03f2e5e633473b8ec5180f44a24919e1MD54THUMBNAILDISSERTAÇÃO Felipe Nunes Walmsley.pdf.jpgDISSERTAÇÃO Felipe Nunes Walmsley.pdf.jpgGenerated Thumbnailimage/jpeg1234https://repositorio.ufpe.br/bitstream/123456789/37647/5/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf.jpg9fdd606b4cadc0949efa572fa952d340MD55CC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-8811https://repositorio.ufpe.br/bitstream/123456789/37647/2/license_rdfe39d27027a6cc9cb039ad269a5db8e34MD52123456789/376472020-08-15 02:10:16.715oai:repositorio.ufpe.br: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ório InstitucionalPUBhttps://repositorio.ufpe.br/oai/requestattena@ufpe.bropendoar:22212020-08-15T05:10:16Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE)false |
dc.title.pt_BR.fl_str_mv |
An investigation into the effects of label noise on dynamic selection algorithms |
title |
An investigation into the effects of label noise on dynamic selection algorithms |
spellingShingle |
An investigation into the effects of label noise on dynamic selection algorithms WALMSLEY, Felipe Nunes Métodos de Ensemble Sistemas de múltiplos classificadores Seleção dinâmica Ruído de classe |
title_short |
An investigation into the effects of label noise on dynamic selection algorithms |
title_full |
An investigation into the effects of label noise on dynamic selection algorithms |
title_fullStr |
An investigation into the effects of label noise on dynamic selection algorithms |
title_full_unstemmed |
An investigation into the effects of label noise on dynamic selection algorithms |
title_sort |
An investigation into the effects of label noise on dynamic selection algorithms |
author |
WALMSLEY, Felipe Nunes |
author_facet |
WALMSLEY, Felipe Nunes |
author_role |
author |
dc.contributor.authorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/8652242028413094 |
dc.contributor.advisorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/8577312109146354 |
dc.contributor.advisor-coLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/6269525393139517 |
dc.contributor.author.fl_str_mv |
WALMSLEY, Felipe Nunes |
dc.contributor.advisor1.fl_str_mv |
CAVALCANTI, George Darmiton da Cunha |
dc.contributor.advisor-co1.fl_str_mv |
SABOURIN, Robert |
contributor_str_mv |
CAVALCANTI, George Darmiton da Cunha SABOURIN, Robert |
dc.subject.por.fl_str_mv |
Métodos de Ensemble Sistemas de múltiplos classificadores Seleção dinâmica Ruído de classe |
topic |
Métodos de Ensemble Sistemas de múltiplos classificadores Seleção dinâmica Ruído de classe |
description |
In the literature on classification problems, it is widely discussed how the presence of label noise can bring about severe degradation in performance. Several works have applied Prototype Selection techniques, Ensemble Methods, or both, in an attempt to alleviate this issue. Nevertheless, these methods are not always able to sufficiently counteract the effects of noise. In this work, we investigate the effects of noise on a particular class of Ensemble Methods, that of Dynamic Selection algorithms, and we are especially interested in the behavior of the Fire-DES++ algorithm, a state of the art algorithm which applies the ENN to algorithm to deal with the effects of noise and imbalance. We propose a method which employs multiple Dynamic Selection sets, based on the Bagging-IH algorithm, which we dub Multiple-Set Dynamic Selection (MSDS), in an attempt to supplant the ENN algorithm on the filtering step. We find that almost all methods based on Dynamic Selection are severely affected by the presence of label noise, with the exception of the KNORAU algorithm. We also find that our proposed method can alleviate the issues caused by noise in some specific scenarios. |
publishDate |
2020 |
dc.date.accessioned.fl_str_mv |
2020-08-14T17:04:59Z |
dc.date.available.fl_str_mv |
2020-08-14T17:04:59Z |
dc.date.issued.fl_str_mv |
2020-01-22 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.citation.fl_str_mv |
WALMSLEY, Felipe Nunes. An investigation into the effects of label noise on dynamic selection algorithms. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020. |
dc.identifier.uri.fl_str_mv |
https://repositorio.ufpe.br/handle/123456789/37647 |
identifier_str_mv |
WALMSLEY, Felipe Nunes. An investigation into the effects of label noise on dynamic selection algorithms. 2020. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Pernambuco, Recife, 2020. |
url |
https://repositorio.ufpe.br/handle/123456789/37647 |
dc.language.iso.fl_str_mv |
por |
language |
por |
dc.rights.driver.fl_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ info:eu-repo/semantics/embargoedAccess |
rights_invalid_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ |
eu_rights_str_mv |
embargoedAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.publisher.program.fl_str_mv |
Programa de Pos Graduacao em Ciencia da Computacao |
dc.publisher.initials.fl_str_mv |
UFPE |
dc.publisher.country.fl_str_mv |
Brasil |
publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFPE instname:Universidade Federal de Pernambuco (UFPE) instacron:UFPE |
instname_str |
Universidade Federal de Pernambuco (UFPE) |
instacron_str |
UFPE |
institution |
UFPE |
reponame_str |
Repositório Institucional da UFPE |
collection |
Repositório Institucional da UFPE |
bitstream.url.fl_str_mv |
https://repositorio.ufpe.br/bitstream/123456789/37647/1/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf https://repositorio.ufpe.br/bitstream/123456789/37647/3/license.txt https://repositorio.ufpe.br/bitstream/123456789/37647/4/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf.txt https://repositorio.ufpe.br/bitstream/123456789/37647/5/DISSERTA%c3%87%c3%83O%20Felipe%20Nunes%20Walmsley.pdf.jpg https://repositorio.ufpe.br/bitstream/123456789/37647/2/license_rdf |
bitstream.checksum.fl_str_mv |
0b7d873c3459043b8269bc0c0ea1e2d0 bd573a5ca8288eb7272482765f819534 03f2e5e633473b8ec5180f44a24919e1 9fdd606b4cadc0949efa572fa952d340 e39d27027a6cc9cb039ad269a5db8e34 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE) |
repository.mail.fl_str_mv |
attena@ufpe.br |
_version_ |
1802310721087733760 |