A Novel Learning Method for ANFIS Using EM Algorithm and Emotional Learning

It is very difficult for the adaptive neuro-fuzzy interference system (ANFIS) using conventional training methods to converge while the samples space distribution is more complex, the desired results for that couldn´t be achieved. To change the situation and improve the learning behavior of ANFIS, in this paper we propose a new self-adaptive learning algorithm for ANFIS differently from conventional training methods. The method firstly adopts the EM algorithm to learning fuzzy parameters of the ANFIS, and then applies emotion learning to learn the Takagi-Sugeno-Kang (TSK) parameters of the linear TSK functions of the ANFIS. The relevant researches indicate that the proposed learning method possesses faster training speed and better adaptability, and is more ubiquitous. In the end, a simulation example shows the availability of the proposed method.