On the behavior of convolutional nets for feature extraction

Deep neural networks are representation learning techniques. During training, a deep net is capable of generating a descriptive language of unprecedented size and detail in machine learning. Extracting the descriptive language coded within a trained CNN model (in the case of image data), and reusing...

Descripción completa

Detalles Bibliográficos
Autores: Garcia-Gasulla, Dario, Parés Pont, Ferran, Vilalta Arias, Armand, Moreno, Jonatan, Ayguadé Parra, Eduard|||0000-0002-5146-103X, Labarta Mancho, Jesús José|||0000-0002-7489-4727, Cortés García, Claudio Ulises|||0000-0003-0192-3096, Suzumura, Toyotaro
Tipo de recurso: artículo
Fecha de publicación:2018
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/115533
Acceso en línea:https://hdl.handle.net/2117/115533
https://dx.doi.org/10.1613/jair.5756
Access Level:acceso abierto
Palabra clave:Neural networks (Computer science)
Artificial intelligence
Deep neural networks
Descriptive language
Machine learning
Deep learning networks
Xarxes neuronals (Informàtica)
Intel·ligència artificial
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
id ES_6b0f4c413e5ccce60eb39cb24c2e9b8a
oai_identifier_str oai:upcommons.upc.edu:2117/115533
network_acronym_str ES
network_name_str España
repository_id_str
spelling On the behavior of convolutional nets for feature extractionGarcia-Gasulla, DarioParés Pont, FerranVilalta Arias, ArmandMoreno, JonatanAyguadé Parra, Eduard|||0000-0002-5146-103XLabarta Mancho, Jesús José|||0000-0002-7489-4727Cortés García, Claudio Ulises|||0000-0003-0192-3096Suzumura, ToyotaroNeural networks (Computer science)Artificial intelligenceDeep neural networksDescriptive languageMachine learningDeep learning networksXarxes neuronals (Informàtica)Intel·ligència artificialÀrees temàtiques de la UPC::Informàtica::Intel·ligència artificialDeep neural networks are representation learning techniques. During training, a deep net is capable of generating a descriptive language of unprecedented size and detail in machine learning. Extracting the descriptive language coded within a trained CNN model (in the case of image data), and reusing it for other purposes is a field of interest, as it provides access to the visual descriptors previously learnt by the CNN after processing millions of images, without requiring an expensive training phase. Contributions to this field (commonly known as feature representation transfer or transfer learning) have been purely empirical so far, extracting all CNN features from a single layer close to the output and testing their performance by feeding them to a classifier. This approach has provided consistent results, although its relevance is limited to classification tasks. In a completely different approach, in this paper we statistically measure the discriminative power of every single feature found within a deep CNN, when used for characterizing every class of 11 datasets. We seek to provide new insights into the behavior of CNN features, particularly the ones from convolutional layers, as this can be relevant for their application to knowledge representation and reasoning. Our results confirm that low and middle level features may behave differently to high level features, but only under certain conditions. We find that all CNN features can be used for knowledge representation purposes both by their presence or by their absence, doubling the information a single CNN feature may provide. We also study how much noise these features may include, and propose a thresholding approach to discard most of it. All these insights have a direct application to the generation of CNN embedding spaces.This work is partially supported by the Joint Study Agreement no. W156463 under the IBM/BSC Deep Learning Center agreement, by the Spanish Government through Programa Severo Ochoa (SEV-2015-0493), by the Spanish Ministry of Science and Technology through TIN2015-65316-P project and by the Generalitat de Catalunya (contracts 2014-SGR-1051), and by the Core Research for Evolutional Science and Technology (CREST) program of Japan Science and Technology Agency (JST).Peer Reviewed20182018-03-0120182018-03-21journal articlehttp://purl.org/coar/resource_type/c_6501VoRhttp://purl.org/coar/version/c_970fb48d4fbd8a85info:eu-repo/semantics/articleapplication/pdfhttps://hdl.handle.net/2117/115533https://dx.doi.org/10.1613/jair.5756reponame:UPCommons. Portal del coneixement obert de la UPCinstname:Universitat Politècnica de Catalunya (UPC)InglésengMinisterio de Economía y Competitividad http://doi.org/10.13039/501100003329 TIN2015-65316-P COMPUTACION DE ALTAS PRESTACIONES VIIopen accesshttp://purl.org/coar/access_right/c_abf2Attribution-NonCommercial-NoDerivs 3.0 Spainhttp://creativecommons.org/licenses/by-nc-nd/3.0/es/info:eu-repo/semantics/openAccessoai:upcommons.upc.edu:2117/1155332026-05-27T15:37:01Z
dc.title.none.fl_str_mv On the behavior of convolutional nets for feature extraction
title On the behavior of convolutional nets for feature extraction
spellingShingle On the behavior of convolutional nets for feature extraction
Garcia-Gasulla, Dario
Neural networks (Computer science)
Artificial intelligence
Deep neural networks
Descriptive language
Machine learning
Deep learning networks
Xarxes neuronals (Informàtica)
Intel·ligència artificial
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
title_short On the behavior of convolutional nets for feature extraction
title_full On the behavior of convolutional nets for feature extraction
title_fullStr On the behavior of convolutional nets for feature extraction
title_full_unstemmed On the behavior of convolutional nets for feature extraction
title_sort On the behavior of convolutional nets for feature extraction
dc.creator.none.fl_str_mv Garcia-Gasulla, Dario
Parés Pont, Ferran
Vilalta Arias, Armand
Moreno, Jonatan
Ayguadé Parra, Eduard|||0000-0002-5146-103X
Labarta Mancho, Jesús José|||0000-0002-7489-4727
Cortés García, Claudio Ulises|||0000-0003-0192-3096
Suzumura, Toyotaro
author Garcia-Gasulla, Dario
author_facet Garcia-Gasulla, Dario
Parés Pont, Ferran
Vilalta Arias, Armand
Moreno, Jonatan
Ayguadé Parra, Eduard|||0000-0002-5146-103X
Labarta Mancho, Jesús José|||0000-0002-7489-4727
Cortés García, Claudio Ulises|||0000-0003-0192-3096
Suzumura, Toyotaro
author_role author
author2 Parés Pont, Ferran
Vilalta Arias, Armand
Moreno, Jonatan
Ayguadé Parra, Eduard|||0000-0002-5146-103X
Labarta Mancho, Jesús José|||0000-0002-7489-4727
Cortés García, Claudio Ulises|||0000-0003-0192-3096
Suzumura, Toyotaro
author2_role author
author
author
author
author
author
author
dc.subject.none.fl_str_mv Neural networks (Computer science)
Artificial intelligence
Deep neural networks
Descriptive language
Machine learning
Deep learning networks
Xarxes neuronals (Informàtica)
Intel·ligència artificial
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
topic Neural networks (Computer science)
Artificial intelligence
Deep neural networks
Descriptive language
Machine learning
Deep learning networks
Xarxes neuronals (Informàtica)
Intel·ligència artificial
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
description Deep neural networks are representation learning techniques. During training, a deep net is capable of generating a descriptive language of unprecedented size and detail in machine learning. Extracting the descriptive language coded within a trained CNN model (in the case of image data), and reusing it for other purposes is a field of interest, as it provides access to the visual descriptors previously learnt by the CNN after processing millions of images, without requiring an expensive training phase. Contributions to this field (commonly known as feature representation transfer or transfer learning) have been purely empirical so far, extracting all CNN features from a single layer close to the output and testing their performance by feeding them to a classifier. This approach has provided consistent results, although its relevance is limited to classification tasks. In a completely different approach, in this paper we statistically measure the discriminative power of every single feature found within a deep CNN, when used for characterizing every class of 11 datasets. We seek to provide new insights into the behavior of CNN features, particularly the ones from convolutional layers, as this can be relevant for their application to knowledge representation and reasoning. Our results confirm that low and middle level features may behave differently to high level features, but only under certain conditions. We find that all CNN features can be used for knowledge representation purposes both by their presence or by their absence, doubling the information a single CNN feature may provide. We also study how much noise these features may include, and propose a thresholding approach to discard most of it. All these insights have a direct application to the generation of CNN embedding spaces.
publishDate 2018
dc.date.none.fl_str_mv 2018
2018-03-01
2018
2018-03-21
dc.type.none.fl_str_mv journal article
http://purl.org/coar/resource_type/c_6501
VoR
http://purl.org/coar/version/c_970fb48d4fbd8a85
dc.type.openaire.fl_str_mv info:eu-repo/semantics/article
format article
dc.identifier.none.fl_str_mv https://hdl.handle.net/2117/115533
https://dx.doi.org/10.1613/jair.5756
url https://hdl.handle.net/2117/115533
https://dx.doi.org/10.1613/jair.5756
dc.language.none.fl_str_mv Inglés
eng
language_invalid_str_mv Inglés
language eng
dc.relation.none.fl_str_mv Ministerio de Economía y Competitividad http://doi.org/10.13039/501100003329 TIN2015-65316-P COMPUTACION DE ALTAS PRESTACIONES VII
dc.rights.none.fl_str_mv open access
http://purl.org/coar/access_right/c_abf2
Attribution-NonCommercial-NoDerivs 3.0 Spain
http://creativecommons.org/licenses/by-nc-nd/3.0/es/
dc.rights.openaire.fl_str_mv info:eu-repo/semantics/openAccess
rights_invalid_str_mv open access
http://purl.org/coar/access_right/c_abf2
Attribution-NonCommercial-NoDerivs 3.0 Spain
http://creativecommons.org/licenses/by-nc-nd/3.0/es/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.source.none.fl_str_mv reponame:UPCommons. Portal del coneixement obert de la UPC
instname:Universitat Politècnica de Catalunya (UPC)
instname_str Universitat Politècnica de Catalunya (UPC)
reponame_str UPCommons. Portal del coneixement obert de la UPC
collection UPCommons. Portal del coneixement obert de la UPC
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869410157160562688
score 15,300719

Similares en LA Referencia