Método AGCRN-CR para Busca Automática de Redes Neurais Artificiais

Anderson Paulo da SilvaTeresa LudermirLeandro Maciel Almeida

This paper describes an automatic search method, CGANN-DC (Cellular Genetic Algorithm + Neural Networks - Direct enCod), for Artificial Neural Networks (ANNs) using Cellular Genetic Algorithms (CGAs). The mais difference between CGAs and Genetic Algorithms (GAs) is the use of a cellular automaton (CA) capable of providing the location for individuals, reducing the possibility of local minima in search space. The CGANN-DC performs evolutionary searches for initial weights, transfer functions, architectures and learning rules. Experimental results have shown that CGANN-DC can find compact networks with a satisfactory generalization power.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: