Variants of memetic and hybrid learning of perceptron networks

Hybrid models combining neural networks and genetic algorithms have been studied recently in order to achieve better performance and/or faster training. In this paper we deal with variants of memetic genetic learning applied for the structure optimization and weights evolution of multilayer perceptron networks. The memetic approach combines genotype and phenotype evolution together with local search represented here by gradient based optimization. It is shown, that combining memetic algorithms with neural networks can lead to better results than relying on neural networks alone in terms of the quality of the solution (both training and generalization error).