On model selection and the disability of neural networks to decompose tasks

Simple item page
dc.creatorToussaint, Marc
dc.date2002-02-19
dc.date.accessioned2026-07-07T05:33:55Z
dc.date.available2026-07-07T05:33:55Z
dc.descriptionA neural network with fixed topology can be regarded as a parametrization of functions, which decides on the correlations between functional variations when parameters are adapted. We propose an analysis, based on a differential geometry point of view, that allows to calculate these correlations. In practise, this describes how one response is unlearned while another is trained. Concerning conventional feed-forward neural networks we find that they generically introduce strong correlations, are predisposed to forgetting, and inappropriate for task decomposition. Perspectives to solve these problems are discussed.
dc.descriptionLaTeX, 7 pages, 3 figures
dc.identifierhttps://arxiv.org/abs/nlin/0202038
dc.identifierhttp://arxiv.org/abs/nlin/0202038
dc.identifierProceedings of the International Joint Conference on Neural Networks (IJCNN 2002), 245-250.
dc.identifier.urihttp://salesiana.dossiersoluciones.com/handle/123456789/80172
dc.subjectAdaptation and Self-Organizing Systems
dc.subjectDisordered Systems and Neural Networks
dc.subjectNeural and Evolutionary Computing
dc.titleOn model selection and the disability of neural networks to decompose tasks
dc.typetext

Files

Collections

arXiv