Genetic procedure for optimization of RBF neural network center positioning [articol]

Abstract

According to RBF neural network theory, it is well known that the (recognition) performances of these architectures depend a lot by the positioning method of centers into input dataspace. Using the affiliation of genetic algorithms to the class of the global searching techniques, and their ability to offer very good results in solving of complex optimization problems, it is justified the attempt to select RBF neural network centers through a suitable genetic procedure.