SIRMs connected fuzzy inference model tuning using genetic algorithm

The single input rule modules (SIRMs) connected inference model is a fuzzy inference model in which a single input rule module is constructed for each system input variable. The output of the module is weighted by the degree of importance for each input and then summarized it into the system output. A tuning algorithm for this model applied to function recognition is suggested based on the steepest descent method. However, the number of rules can not be optimized. In this work, a tuning process based on the genetic algorithm is proposed. It allows a wide search for tuned parameters with optimized number of rules. A nonlinear function recognition simulation experiment is done to confirm the validity of the proposed method.