Hierarchical fuzzy reasoning: adaptive structure and rule by genetic algorithms

This paper proposes a self-tuning hierarchical fuzzy reasoning that uses the Genetic Algorithm and back-propagation method. If a fuzzy system has a number of inputs, the number of membership functions and rules will be exploded. Therefore, it is necessary to reduce the number of membership functions. One method to do so is to make a hierarchical structure of fuzzy inference units that have a few inputs. However the hierarchical structure cannot be made without considering the relationship among inputs. The proposed method is based on the Genetic Algorithm with a strategy that favors systems with fewer rules and membership functions, and obtains the optimal structure. The proposed method is applied to multi-dimensional function approximation problems in order to show the effectiveness