The method of non-linear learning the neuro-fuzzy inference system

The new method of learning the neuro-fuzzy inference system is considered in the article. A feature of this method is the use of a non-linear function in the model of the differential areas. The generalized model of fuzzy inference whit using the linear and nonlinear functions is structured. The feature of the generalized model is use of various t-norms (Mamdani implication, the implication of the algebraic product, Lukasevic's implication, the implication of the bounded difference, and soft implication). The results of modeling the learning process of the neuro-fuzzy inference system using the linear and non-linear functions in the method of the differential areas are presented. Evaluation of the working of the neurofuzzy inference system was carried out on the basis of calculating the RMSE and the time required for its learning. The proposed nonlinear model of the differential areas increases the accuracy during learning fuzzy systems. This conclusion is confirmed by the results of simulation modeling presented in the article.