Design of Fuzzy Neural Network for Function Approximation and Classi¯cation

A hybrid Fuzzy Neural Network (FNN) system is presented in this paper. The proposed FNN can handle numeric and fuzzy inputs simultaneously. The numeric inputs are fuzzified by input nodes upon presentation to the network while the fuzzy inputs do not require this translation. The connections between input to hidden nodes represent rule antecedents and hidden to output nodes represent rule consequents. All the connections are represented by Gaussian fuzzy sets. The mutual subsethood measure for fuzzy sets that indicates the degree to which the two fuzzy sets are equal and is used as a method of activation spread in the network. A volume based defuzzification method is used to compute the numeric output of the network. The training of the network is done using gradient descent learning procedure. The model has been tested on three benchmark problems i.e. sine—cosine and Narazaki Ralescuʹs function for approximation and Iris flower data for classification. Results are also compared with existing schemes and the proposed model shows its natural capability as a function approximator, and classifier.