Simplification of fuzzy-neural systems using similarity analysis

A fuzzy neural network system (FNSS) is proposed for implementing fuzzy inference systems for system modeling. The use of fuzzy similarity measure helps to eliminate redundant fuzzy rules. This reduces the complexity of the network and makes it suitable for industrial applications. Computer simulation studies indicate a satisfactory performance of FNSS. A novel and efficient on-line initialization method is also designed for selecting the initial parameters of the FNNS.