Abstract
In this study, we introduce a class of neural architectures of self-organizing neural networks (SONN) that is based on a genetically optimized multilayer perceptron with polynomial neurons (PNs) or fuzzy polynomial neurons (FPNs). We discuss its comprehensive design methodology involving mechanisms of genetic optimization, especially genetic algorithms (GAs). The conventional SONN is based on some mechanisms of self-organization and an evolutionary algorithm rooted in the extended group method of data handling (GMDH) method. In contrast, the proposed genetically optimized SONN (called “gSONN”, for brief) results in a structurally optimized structure and comes with a higher level of flexibility in comparison to the one encountered in the conventional SONN. This structural optimization is realized via GAs whereas in the case of the parametric optimization we proceed with a standard least square method-based learning. Through the consecutive process of such structural and parametric optimization, the gSONN becomes generated in a highly dynamic fashion. The performance of the network is quantified through experimentation in which we exploit standard data already used in fuzzy or neurofuzzy modeling. These results highlight the superiority of the proposed networks over the existing fuzzy and neural models.