Structure and learning algorithms of a nonsingleton input fuzzy neural network based on the approximate analogical reasoning schema

摘要

A novel fuzzy neural network, the pseudo-outer-product-based fuzzy neural network using the nonsingleton fuzzifier together with the approximate analogical reasoning schema (nonsingleton fuzzifier POPFNN-AARS), is proposed in this paper. The nonsingleton fuzzifier POPFNN-AARS is developed as an improvement over the singleton fuzzifier POPFNN-AARS developed by Quek and Zhou [POPFNN-AARS: a pseudo outer-product based fuzzy neural network, IEEE Trans. Systems, Man, Cybernetics 29(6) (1999) 859-870]. The employment of the approximate analogical reasoning schema (AARS) as the fuzzy inference model enables the integration of a nonsingleton fuzzifier in the nonsingleton fuzzifier POPFNN-AARS. Compared against the commonly used singleton fuzzifier, the nonsingleton fuzzifier is more powerful in handling imprecise data. Different similarity measures and modification functions proposed by Zwick et al. [Measures of similarity between fuzzy concepts: a comparative analysis, Internal. J. Approx. Reason. 1 (1987) 221-242] for AARS in the nonsingleton fuzzifier-AARS are investigated. The structure and learning algorithms of the proposed nonsingleton fuzzifier POPFNN-AARS are presented. Several sets of real-life data are used to test the performance of the nonsingleton fuzzifier POPFNN-AARS and their experimental results are presented for detailed discussion.