A hybrid evolutionary design of neuro-fuzzy systems

In this paper, a hybrid evolutionary approach, combining the theory of learning automata (LA) and the steady-state genetic algorithm (SSGA), is proposed for design of TSKtype fuzzy model (TFM). In the proposed memetic approach, both the number of fuzzy rules and adjustable parameters in the TFM are designed concurrently. A learning automaton, which systematically updates a strategy to enhance the performance in response to the output results, is used to find the optimal number of rules, whereas the SSGA is used to perform the tuning of the TFM parameters. Computer simulations have demonstrated that the proposed hybrid method performs better than some existing methods.