FUZZY INFERENCE MODELING BASED ON FUZZY SINGLETON-TYPE REASONING

Abstract. In this paper, the idea of the neuro-fuzzy learning algorithm has been extended, by which the tuning parameters in the fuzzy rules can be learned without changing the fuzzy rule table form used in usual fuzzy applications. A new neuro-fuzzy learning algorithm in the case of the fuzzy singleton-type reasoning method has been proposed. Due to the flexibility of the fuzzy singleton-type reasoning method, the extended method is more reasonable and suitable for constructing an optimum fuzzy system model than the conventional neuro-fuzzy learning algorithm. Moreover, the efficiency of the extended neuro-fuzzy learning algorithm compared to a genetic algorithm is demonstrated by identifying a nonlinear function. Keywords: Fuzzy singleton-type reasoning, Fuzzy rule table, Triangular-type membership function, Matching approach, Neuro-fuzzy learning algorithm