Learning TSK Fuzzy Model by GA-BP Method

It is difficult to learn TSK fuzzy model because the problem is multiconstraint and multitarget optimization. GA-BP hybrid learning method for the model is proposed. Some problems related to a species coding means for the model structure, evolution and fitness evaluation strategy are discussed. The error back propagation algorithm (BP) for training the antecedent and consequent parameters during the process of evolution is inferred. The characteristic of the method requests a little of previous information about objects, and avoids slow convergence, and has better adaptive capability, and is able to obtain compact and accurate fuzzy model from samples, the validity of the method has been demonstrated by an example of function approximation.