Múltiplas Redes Neurais Utilizando Redes MLP e RBF

Abstract

This work describes a methodology that promotes a collaboration and selfcoordination between Multilayer Perceptron and Radial Base Function networks through the Multiple Neural Networks technique, which doesn't require the use of a coordinator. Multiple Neural Networks are used for the purpose of facilitating learning and solving very complex problems. Normally, in order to learn an unimpaired network or learn from others, use a coordinator, whose function is to avoid conflicts between networks. However, a technique has been developed that dispenses with the use of the element coordinator, where as self-coordinating networks. In order to extend the scope of this technique, this work proposes a collaboration between networks of different types, making their advantages collaborate in order to efficiently solve diverse problems. This methodology is compared to other solutions for detection of breast cancer and disorders of the spine.