HYBRID NEURON NETWORKS BASED ON Q- , W- AND CLASSICAL NEURONS

Abstract

The problem of structural-parametric synthesis of hybrid neural network based on the use of multilayer perceptron topology is considered. Hybridization is achieved through the use of artificial neurons of different types, namely Q-neuron, W-neuron and classical neuron. The problem of optimal selection of the number of layers, neurons in layers, as well as the types of neurons in each layer and the principles of alternating them using the genetic algorithm SPEA2 is solved. Examples of building a hybrid neural network using this methodology and a given optimization criterion for solving classification and forecasting problems are given.