Extracting of Specific Knowledge through RBF-Fuzzy Networks and Genetic Algorithms

This paper proposes the application of Genetic Learning as a procedure for the optimal design and training of neuro-fuzzy systems. Once this learning procedure has been implemented, hybridization between Genetic Algorithms (GA) and the traditional local search technique is carried out to form a Hybrid Algorithm, in order to achieve the maximum possible efficiency in the search, and to be able to exceed the performance of traditional techniques. The proposed neuro-fuzzy model and learning methods have some restrictions in order to find coherent fuzzy rules of the problem to be solved. Effectiveness of the developed techniques when solving the problem of identification of functions will be assessed in comparison to that of habitual training techniques, with the help of many practical tests. Finally, conclusions will be drawn in terms of the test results. Key-Words: Fuzzy Logic, Neural Networks, Genetic Algorithms, Neuro-Fuzzy Models, Radial Basis Functions (RBF), Supervised Learning Algorithm.