X-GAN: Generative Adversarial Networks Training Guided with Explainable Artificial Intelligence

Registro completo de metadados
MetadadosDescriçãoIdioma
Autor(es): dc.contributorUniversity of Bologna-
Autor(es): dc.contributorUniversity of Porto (FEUP)-
Autor(es): dc.contributorUniversidade de São Paulo (USP)-
Autor(es): dc.contributorUniversidade Federal de Uberlândia (UFU)-
Autor(es): dc.contributorUniversidade Estadual Paulista (UNESP)-
Autor(es): dc.creatorRozendo, Guilherme Botazzo-
Autor(es): dc.creatorLumini, Alessandra-
Autor(es): dc.creatorRoberto, Guilherme Freire-
Autor(es): dc.creatorTosta, Thaína Aparecida Azevedo-
Autor(es): dc.creatordo Nascimento, Marcelo Zanchetta-
Autor(es): dc.creatorNeves, Leandro Alves-
Data de aceite: dc.date.accessioned2025-08-21T19:14:50Z-
Data de disponibilização: dc.date.available2025-08-21T19:14:50Z-
Data de envio: dc.date.issued2025-04-29-
Data de envio: dc.date.issued2023-12-31-
Fonte completa do material: dc.identifierhttp://dx.doi.org/10.5220/0012618400003690-
Fonte completa do material: dc.identifierhttps://hdl.handle.net/11449/306242-
Fonte: dc.identifier.urihttp://educapes.capes.gov.br/handle/11449/306242-
Descrição: dc.descriptionGenerative Adversarial Networks (GANs) create artificial images through adversary training between a generator (G) and a discriminator (D) network. This training is based on game theory and aims to reach an equilibrium between the networks. However, this equilibrium is hardly achieved, and D tends to be more powerful. This problem occurs because G is trained based on only a single value representing D’s prediction, and only D has access to the image features. To address this issue, we introduce a new approach using Explainable Artificial Intelligence (XAI) methods to guide the G training. Our strategy identifies critical image features learned by D and transfers this knowledge to G. We have modified the loss function to propagate a matrix of XAI explanations instead of only a single error value. We show through quantitative analysis that our approach can enrich the training and promote improved quality and more variability in the artificial images. For instance, it was possible to obtain an increase of up to 37.8% in the quality of the artificial images from the MNIST dataset, with up to 4.94% more variability when compared to traditional methods.-
Descrição: dc.descriptionFundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)-
Descrição: dc.descriptionCoordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)-
Descrição: dc.descriptionFundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG)-
Descrição: dc.descriptionDepartment of Computer Science and Engineering (DISI) University of Bologna-
Descrição: dc.descriptionFaculty of Engineering University of Porto (FEUP)-
Descrição: dc.descriptionScience and Technology Institute (ICT) Federal University of São Paulo (UNIFESP)-
Descrição: dc.descriptionFaculty of Computer Science (FACOM) Federal University of Uberlândia (UFU)-
Descrição: dc.descriptionDepartment of Computer Science and Statistics (DCCE) São Paulo State University-
Descrição: dc.descriptionDepartment of Computer Science and Statistics (DCCE) São Paulo State University-
Descrição: dc.descriptionFAPESP: #2022/03020-1-
Descrição: dc.descriptionCAPES: #311404/2021-9-
Descrição: dc.descriptionCAPES: #313643/2021-0-
Descrição: dc.descriptionFAPEMIG: #APQ-00578-18-
Formato: dc.format674-681-
Idioma: dc.languageen-
Relação: dc.relationInternational Conference on Enterprise Information Systems, ICEIS - Proceedings-
???dc.source???: dc.sourceScopus-
Palavras-chave: dc.subjectExplainable Artificial Intelligence-
Palavras-chave: dc.subjectGAN Training-
Palavras-chave: dc.subjectGenerative Adversarial Networks-
Título: dc.titleX-GAN: Generative Adversarial Networks Training Guided with Explainable Artificial Intelligence-
Tipo de arquivo: dc.typeaula digital-
Aparece nas coleções:Repositório Institucional - Unesp

Não existem arquivos associados a este item.