Successive Overrelaxation for Mamdani Fuzzy Systems

To design a Mamdani fuzzy system with good generalization ability in high dimensional feature space, a novel learning algorithm based on the structural risk minimization (SRM) inductive principle is presented in this paper. Firstly, the parameter estimation of a Mamdani fuzzy system is converted to a quadratic optimization problem. Then, a versatile iterative method, successive overrelaxation, is proposed. In the proposed algorithm, the fuzzy kernel generated by premise membership functions is proved to be a Mercer kernel. Numerical experiments show that the presented algorithm improves the generalization ability of Mamdani fuzzy systems.