Hybrid Training of Morphological Neural Networks Applied to Function Approximation

André Luís Vasconcelos CoelhoClodoaldo Aparecido de Moraes LimaMário Ernesto de Souza e SilvaFernando J. Von Zuben

In this paper, we present a hybrid algorithm for training a morphological neural network, which combines an evolutionary programming technique with a non-linear optimization method based on gradient information. The aim behind such fusion of techniques is to properly exploit the high non-linearity features exhibited by the morphological neuron. The presented simulation results seek to demonstrate the viability of applying this new training algorithm to function approximation and classification problems, comparing its performance with the multi-layer perceptron in complex functions. Clique no link abaixo para buscar o texto completo deste trabalho na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: