Fuzzy reasoning is the important part of fuzzy logic. The reductive property is so far the only criterion to illustrate if a fuzzy reasoning method is good.In this paper, first a definition is proposed of the reductive property of fuzzy reasoning with multiple rules, and then research on the existing fuzzy reasoning methods. Zadeh's CRI method is applied in fuzzy reasoning wildly, but it is not reductive. We first give examples to show that Wang Guojun's Triple I Method, the Truth-valued Flow Inference and other similarity-based fuzzy reasoning methods do not have the reductive property for multi-rules fuzzy inference.After research on the different definitions of similarity measures and several similarity-based fuzzy reasoning methods, an algorithm of general fuzzy inference with the reductive property is proposed in this paper. It is named CON-FI algorithm. Finally, the simulation results show that the algorithm of general fuzzy inference with the reductive property is a good fuzzy reasoning method.
|