Learning to weight similarity measures with Siamese networks: A case study on optimum-path forest

Registro completo de metadados
MetadadosDescriçãoIdioma
Autor(es): dc.contributorUniversidade Estadual Paulista (UNESP)-
Autor(es): dc.creatorDe Rosa, Gustavo H.-
Autor(es): dc.creatorPapa, João Paulo-
Data de aceite: dc.date.accessioned2025-08-21T22:46:09Z-
Data de disponibilização: dc.date.available2025-08-21T22:46:09Z-
Data de envio: dc.date.issued2023-03-01-
Data de envio: dc.date.issued2023-03-01-
Data de envio: dc.date.issued2022-01-23-
Fonte completa do material: dc.identifierhttp://dx.doi.org/10.1016/B978-0-12-822688-9.00015-3-
Fonte completa do material: dc.identifierhttp://hdl.handle.net/11449/241424-
Fonte: dc.identifier.urihttp://educapes.capes.gov.br/handle/11449/241424-
Descrição: dc.descriptionRecent advances in machine learning algorithms have been aiding humans and improving their decision-making capacities in various applications, such as medical imaging, image classification and reconstruction, object recognition, and text categorization. A graph-based classifier, known as Optimum-Path Forest (OPF), has been extensively researched in the last years, mainly due to its parameterless nature and state-of-the-art results compared to well-known literature classifiers, for example, support vector machines. Nevertheless, one drawback concerning such an approach lies in its distance calculation, which has to be selected from a range of formulae and computed between all nodes to weigh the graph's arcs, and hence time-consuming. Therefore in this work, we propose to address such a problem by precomputing the arcs' distances through a similarity measure obtained from Siamese networks. The idea is to employ the same training set used by the OPF classifier to train a Siamese network and calculate the samples' distance through a similarity measure. The experimental results show that the proposed method is suitable, where the similarity-based OPF achieved comparable results to its standard counterpart and even surpassed it in some datasets. Additionally, the precalculated similarity matrix lessens the burden of recalculating the distances for every new classification. © 2022 Copyright-
Descrição: dc.descriptionDepartment of Computing São Paulo State University, Bauru-
Descrição: dc.descriptionUNESP - São Paulo State University School of Sciences-
Descrição: dc.descriptionDepartment of Computing São Paulo State University, Bauru-
Descrição: dc.descriptionUNESP - São Paulo State University School of Sciences-
Formato: dc.format155-173-
Idioma: dc.languageen-
Relação: dc.relationOptimum-Path Forest: Theory, Algorithms, and Applications-
???dc.source???: dc.sourceScopus-
Palavras-chave: dc.subjectMachine learning-
Palavras-chave: dc.subjectOptimum-path forest-
Palavras-chave: dc.subjectSiamese network-
Palavras-chave: dc.subjectSimilarity function-
Palavras-chave: dc.subjectSupervised classification-
Título: dc.titleLearning to weight similarity measures with Siamese networks: A case study on optimum-path forest-
Tipo de arquivo: dc.typelivro digital-
Aparece nas coleções:Repositório Institucional - Unesp

Não existem arquivos associados a este item.