A New Approach Based on Fuzzy-Adaptive Structure & Parameter Learning Applied in Meta-Cognitive Algorithm for ANFIS

A fuzzy inference system (FIS) is a widely used modeling and simulating approach based on the concept of fuzzy set theory, fuzzy if-then rules and fuzzy logic systems. All fuzzy inference systems have an adaptive composition that parameters are set during the training process. There is not any especial instruction to change the training parameters or the structure of the fuzzy inference systems during the training process. In this paper, a new online approach is proposed to adapt some systematic parameters of a typical fuzzy inference system. Employing this approach, some training and systematic parameters are selected during the training procedure to be constant or to be adaptive. This approach is realized and examined on the Adaptive Neural-fuzzy Inference System (ANFIS) and Meta-cognitive neuro fuzzy inference system (McFIS). The training process of the McFIS consists of two portions which includes a cognitive and a meta-cognitive one. A training mechanism controls the learning process of the cognitive portion. This paper illustrates the meta-cognitive training algorithm for parameterization of the fuzzy inference systems and applies the proposed approach for functional and systematic or structural parameter selection in the meta-cognitive algorithm (McFIS) which ends to some noticeable merits. The idea is examined by numerical examples and the results are compared with other methods. It is shown that the proposed approach reveals acceptable and efficient performance in the structure selection problem.