Mostra el registre d'ítem simple
Feature discriminativity estimation in CNNs for transfer learning
dc.contributor.author | Giménez Ábalos, Víctor |
dc.contributor.author | Vilalta Arias, Armand |
dc.contributor.author | Garcia Gasulla, Dario |
dc.contributor.author | Labarta Mancho, Jesús José |
dc.contributor.author | Ayguadé Parra, Eduard |
dc.contributor.other | Universitat Politècnica de Catalunya. Departament d'Arquitectura de Computadors |
dc.contributor.other | Barcelona Supercomputing Center |
dc.date.accessioned | 2020-05-05T09:15:35Z |
dc.date.available | 2020-05-05T09:15:35Z |
dc.date.issued | 2019 |
dc.identifier.citation | Giménez, V. [et al.]. Feature discriminativity estimation in CNNs for transfer learning. A: International Conference of the Catalan Association for Artificial Intelligence. "Proceedings of the 22nd International Conference of the Catalan Association for Artificial Intelligence". IOS Press, 2019, p. 64-73. |
dc.identifier.isbn | 978-1-64368-015-6 |
dc.identifier.uri | http://hdl.handle.net/2117/186300 |
dc.description.abstract | The purpose of feature extraction on convolutional neural networks is to reuse deep representations learnt for a pre-trained model to solve a new, potentially unrelated problem. However, raw feature extraction from all layers is unfeasible given the massive size of these networks. Recently, a supervised method using complexity reduction was proposed, resulting in significant improvements in performance for transfer learning tasks. This approach first computes the discriminative power of features, and then discretises them using thresholds computed for the task. In this paper, we analyse the behaviour of these thresholds, with the purpose of finding a methodology for their estimation. After a comprehensive study, we find a very strong correlation between problem size and threshold value, with coefficient of determination above 90%. These results allow us to propose a unified model for threshold estimation, with potential application to transfer learning tasks. |
dc.description.sponsorship | This work is partially supported by BSC-IBM Deep Learning Center agreement, the Spanish Government through Programa Severo Ochoa (SEV-2015-0493), the Spanish Ministry of Science and Technology through TIN2015-65316-P project and the Generalitat de Catalunya (contract 2017-SGR-1414). |
dc.format.extent | 10 p. |
dc.language.iso | eng |
dc.publisher | IOS Press |
dc.subject | Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial::Aprenentatge automàtic |
dc.subject.lcsh | Machine learning |
dc.subject.lcsh | Neural networks (Computer science) |
dc.subject.other | Transfer learning |
dc.subject.other | Machine learning |
dc.subject.other | CNN |
dc.subject.other | Feature extraction |
dc.title | Feature discriminativity estimation in CNNs for transfer learning |
dc.type | Conference report |
dc.subject.lemac | Aprenentatge automàtic |
dc.subject.lemac | Xarxes neuronals (Informàtica) |
dc.contributor.group | Universitat Politècnica de Catalunya. CAP - Grup de Computació d'Altes Prestacions |
dc.identifier.doi | 10.3233/FAIA190109 |
dc.description.peerreviewed | Peer Reviewed |
dc.relation.publisherversion | http://ebooks.iospress.nl/volumearticle/52822 |
dc.rights.access | Open Access |
local.identifier.drac | 27852291 |
dc.description.version | Postprint (author's final draft) |
dc.relation.projectid | info:eu-repo/grantAgreement/MINECO//TIN2015-65316-P/ES/COMPUTACION DE ALTAS PRESTACIONES VII/ |
dc.relation.projectid | info:eu-repo/grantAgreement/AGAUR/2017 SGR 1414 |
dc.relation.projectid | info:eu-repo/grantAgreement/MINECO//SEV-2015-0493/ES/BARCELONA SUPERCOMPUTING CENTER - CENTRO. NACIONAL DE SUPERCOMPUTACION/ |
local.citation.author | Giménez, V.; Vilalta, A.; Garcia-Gasulla, D.; Labarta, J.; Ayguadé, E. |
local.citation.contributor | International Conference of the Catalan Association for Artificial Intelligence |
local.citation.publicationName | Proceedings of the 22nd International Conference of the Catalan Association for Artificial Intelligence |
local.citation.startingPage | 64 |
local.citation.endingPage | 73 |