Layer factor analysis in convolutional neural networks for explainability

Explanatory methods that focus on the analysis of the features encoded by Convolutional Neural Networks (CNNs) are of great interest, since they help to understand the underlying process hidden behind the black-box nature of these models. However, to explain the knowledge gathered in a given layer,...

Descripción completa

Detalles Bibliográficos
Autores: López González, Clara Isabel, Gómez Silva, María José, Besada Portas, Eva, Pajares Martínsanz, Gonzalo
Tipo de recurso: artículo
Fecha de publicación:2024
País:España
Institución:Universidad Complutense de Madrid (UCM)
Repositorio:Docta Complutense
Idioma:inglés
OAI Identifier:oai:docta.ucm.es:20.500.14352/98403
Acceso en línea:https://hdl.handle.net/20.500.14352/98403
Access Level:acceso abierto
Palabra clave:004.8
004.85
004.932
004.032.26
Deep learning
Explainable Artificial Intelligence (xAI)
Statistical modeling
Visual explanation
Feature learning
Attribution map
Inteligencia artificial (Informática)
1203.17 Informática
id ES_22cc87defdaece648d569bfe7e20f8ea
oai_identifier_str oai:docta.ucm.es:20.500.14352/98403
network_acronym_str ES
network_name_str España
repository_id_str
spelling Layer factor analysis in convolutional neural networks for explainabilityLópez González, Clara IsabelGómez Silva, María JoséBesada Portas, EvaPajares Martínsanz, Gonzalo004.8004.85004.932004.032.26Deep learningExplainable Artificial Intelligence (xAI)Statistical modelingVisual explanationFeature learningAttribution mapInteligencia artificial (Informática)1203.17 InformáticaExplanatory methods that focus on the analysis of the features encoded by Convolutional Neural Networks (CNNs) are of great interest, since they help to understand the underlying process hidden behind the black-box nature of these models. However, to explain the knowledge gathered in a given layer, they must decide which of the numerous filters to study, further assuming that each of them corresponds to a single feature. This, coupled with the redundancy of information, makes it difficult to ensure that the relevant characteristics are being analyzed. The above represents an important challenge and defines the aim and scope of our proposal. In this paper we present a novel method, named Explainable Layer Factor Analysis for CNNs (ELFA-CNNs), which models and describes with quality convolutional layers relying on factor analysis. Regarding contributions, ELFA obtains the essential underlying features, together with their correlation with the original filters, providing an accurate and well-founded summary. Through the factorial parameters we gain insights about the information learned, the connections between channels, and the redundancy of the layer, among others. To provide visual explanations in a similarly way to other methods, two additional proposals are made: a) Essential Feature Attribution Maps (EFAM) and b) intrinsic features inversion. The results prove the effectiveness of the developed general methods. They are evaluated in different CNNs (VGG-16, ResNet-50, and DeepLabv3+) on generic datasets (CIFAR-10, imagenette, and CamVid). We demonstrate that convolutional layers adequately fit a factorial model thanks to the new metrics presented for factor and fitting residuals (D1, D>, and Res, derive from covariance matrices). Moreover, knowledge about the deep image representations and the learning process is acquired, as well as reliable heat maps highlighting regions where essential features are located. This study effectively provides an explainable approach that can be applied to different CNNs and over different datasets.ElsevierUniversidad Complutense de Madrid20242024-01-0120242024-01-01journal articlehttp://purl.org/coar/resource_type/c_6501VoRhttp://purl.org/coar/version/c_970fb48d4fbd8a85info:eu-repo/semantics/articleapplication/pdfhttps://hdl.handle.net/20.500.14352/98403reponame:Docta Complutenseinstname:Universidad Complutense de Madrid (UCM)InglésengComunidad de Madrid http://dx.doi.org/10.13039/100012818 PRICIT Y2020MCIN AEI 10.13039 EUMCIN PEICTI PID2021-127648OB-C33 Cooperación de vehículos de superficie y aéreos para aplicaciones de inspección en entornos cambiantesopen accesshttp://purl.org/coar/access_right/c_abf2Attribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/info:eu-repo/semantics/openAccessoai:docta.ucm.es:20.500.14352/984032026-06-02T12:44:21Z
dc.title.none.fl_str_mv Layer factor analysis in convolutional neural networks for explainability
title Layer factor analysis in convolutional neural networks for explainability
spellingShingle Layer factor analysis in convolutional neural networks for explainability
López González, Clara Isabel
004.8
004.85
004.932
004.032.26
Deep learning
Explainable Artificial Intelligence (xAI)
Statistical modeling
Visual explanation
Feature learning
Attribution map
Inteligencia artificial (Informática)
1203.17 Informática
title_short Layer factor analysis in convolutional neural networks for explainability
title_full Layer factor analysis in convolutional neural networks for explainability
title_fullStr Layer factor analysis in convolutional neural networks for explainability
title_full_unstemmed Layer factor analysis in convolutional neural networks for explainability
title_sort Layer factor analysis in convolutional neural networks for explainability
dc.creator.none.fl_str_mv López González, Clara Isabel
Gómez Silva, María José
Besada Portas, Eva
Pajares Martínsanz, Gonzalo
author López González, Clara Isabel
author_facet López González, Clara Isabel
Gómez Silva, María José
Besada Portas, Eva
Pajares Martínsanz, Gonzalo
author_role author
author2 Gómez Silva, María José
Besada Portas, Eva
Pajares Martínsanz, Gonzalo
author2_role author
author
author
dc.contributor.none.fl_str_mv Universidad Complutense de Madrid
dc.subject.none.fl_str_mv 004.8
004.85
004.932
004.032.26
Deep learning
Explainable Artificial Intelligence (xAI)
Statistical modeling
Visual explanation
Feature learning
Attribution map
Inteligencia artificial (Informática)
1203.17 Informática
topic 004.8
004.85
004.932
004.032.26
Deep learning
Explainable Artificial Intelligence (xAI)
Statistical modeling
Visual explanation
Feature learning
Attribution map
Inteligencia artificial (Informática)
1203.17 Informática
description Explanatory methods that focus on the analysis of the features encoded by Convolutional Neural Networks (CNNs) are of great interest, since they help to understand the underlying process hidden behind the black-box nature of these models. However, to explain the knowledge gathered in a given layer, they must decide which of the numerous filters to study, further assuming that each of them corresponds to a single feature. This, coupled with the redundancy of information, makes it difficult to ensure that the relevant characteristics are being analyzed. The above represents an important challenge and defines the aim and scope of our proposal. In this paper we present a novel method, named Explainable Layer Factor Analysis for CNNs (ELFA-CNNs), which models and describes with quality convolutional layers relying on factor analysis. Regarding contributions, ELFA obtains the essential underlying features, together with their correlation with the original filters, providing an accurate and well-founded summary. Through the factorial parameters we gain insights about the information learned, the connections between channels, and the redundancy of the layer, among others. To provide visual explanations in a similarly way to other methods, two additional proposals are made: a) Essential Feature Attribution Maps (EFAM) and b) intrinsic features inversion. The results prove the effectiveness of the developed general methods. They are evaluated in different CNNs (VGG-16, ResNet-50, and DeepLabv3+) on generic datasets (CIFAR-10, imagenette, and CamVid). We demonstrate that convolutional layers adequately fit a factorial model thanks to the new metrics presented for factor and fitting residuals (D1, D>, and Res, derive from covariance matrices). Moreover, knowledge about the deep image representations and the learning process is acquired, as well as reliable heat maps highlighting regions where essential features are located. This study effectively provides an explainable approach that can be applied to different CNNs and over different datasets.
publishDate 2024
dc.date.none.fl_str_mv 2024
2024-01-01
2024
2024-01-01
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/20.500.14352/98403
url https://hdl.handle.net/20.500.14352/98403
dc.language.none.fl_str_mv Inglés
eng
language_invalid_str_mv Inglés
language eng
dc.relation.none.fl_str_mv Comunidad de Madrid http://dx.doi.org/10.13039/100012818 PRICIT Y2020
MCIN AEI 10.13039 EU
MCIN PEICTI PID2021-127648OB-C33 Cooperación de vehículos de superficie y aéreos para aplicaciones de inspección en entornos cambiantes
dc.rights.none.fl_str_mv open access
http://purl.org/coar/access_right/c_abf2
Attribution 4.0 International
http://creativecommons.org/licenses/by/4.0/
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 4.0 International
http://creativecommons.org/licenses/by/4.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Elsevier
publisher.none.fl_str_mv Elsevier
dc.source.none.fl_str_mv reponame:Docta Complutense
instname:Universidad Complutense de Madrid (UCM)
instname_str Universidad Complutense de Madrid (UCM)
reponame_str Docta Complutense
collection Docta Complutense
repository.name.fl_str_mv
repository.mail.fl_str_mv
_version_ 1869404602748633088
score 15,300719

Similares en LA Referencia