Título |
Trainable and explainable simplicial map neural networks |
Autores |
PALUZO HIDALGO, EDUARDO, Gonzalez-Diaz R. , Gutiérrez-Naranjo M.A. |
Publicación externa |
Si |
Medio |
INFORMATION SCIENCES |
Alcance |
Article |
Naturaleza |
Científica |
Cuartil JCR |
1 |
Cuartil SJR |
1 |
Web |
https://www.scopus.com/inward/record.uri?eid=2-s2.0-85188534762&doi=10.1016%2fj.ins.2024.120474&partnerID=40&md5=ab15e558c17011ae2ad7566f090fa602 |
Fecha de publicacion |
01/01/2024 |
ISI |
001218226500001 |
Scopus Id |
2-s2.0-85188534762 |
DOI |
10.1016/j.ins.2024.120474 |
Abstract |
Simplicial map neural networks (SMNNs) are topology-based neural networks with interesting properties such as universal approximation ability and robustness to adversarial examples under appropriate conditions. However, SMNNs present some bottlenecks for their possible application in high-dimensional datasets. First, SMNNs have precomputed fixed weight and no SMNN training process has been defined so far, so they lack generalization ability. Second, SMNNs require the construction of a convex polytope surrounding the input dataset. In this paper, we overcome these issues by proposing an SMNN training procedure based on a support subset of the given dataset and replacing the construction of the convex polytope by a method based on projections to a hypersphere. In addition, the explainability capacity of SMNNs and effective implementation are also newly introduced in this paper. © 2024 The Author(s) |
Palabras clave |
Approximation ability; Condition; Convex polytopes; Explainable artificial intelligence; Neural networks trainings; Neural-networks; Property; Simplicial map; Training neural network; Universal approximation |
Miembros de la Universidad Loyola |
|