Local and global estimation of Takagi-Sugeno consequent parameters in genetic fuzzy systems

In modular and hierarchical evolutionary design of Takagi-Sugeno (T-S) fuzzy systems, an important issue involves the determination of an effective procedure to optimize rule consequent parameters. All the aspects associated with the antecedent part of each fuzzy rule are evolved through generations, and given a specification of the antecedent part of the rules that compose a candidate fuzzy system, the best set of consequent parameters should be determined. This paper investigates the use of global and local least squares optimization procedures to perform this task. Function approximation problems are solved to test the performance of the evolutionary process in comparison with alternative solutions.