SELEÇÃO DE MODELOS BASEADO NA APLICAÇÃO DE PCA À REDE MLP DUAL

Título: SELEÇÃO DE MODELOS BASEADO NA APLICAÇÃO DE PCA À REDE MLP DUAL

Autores: Medeiros, Cláudio M. S.; Barreto, Guilherme A.

Resumo: In this paper we present a novel method for pruning redundant hidden neurons of a trained multilayer Perceptron (MLP). The proposed method is based on the principal components analysis (PCA) applied to the dual network, which is linear in parameters. Simulations using real-world data indicate that the proposed method presents equivalent or better performance than traditional pruning techniques.

Palavras-chave: Model Selection; MLP; Dual Network; PCA

Páginas: 5

Código DOI: 10.21528/CBRN2009-057

Artigo em PDF: 057_CBRN2009.pdf

Arquivo BibTex: 057_CBRN2009.bib