Neuro-fuzzy modeling with a new hybrid learning

In this paper, a neuro-fuzzy modeling approach with a new hybrid-learning algorithm (NF-HLA) is presented. The NF-HLA is built using a feed forward neural network functionally equivalent to a Takagi-Sugeno fuzzy system. At the premise part of the NF-HLA, the parameters of the membership functions are adjusted with the use of the Levenberg-Marquardt algorithm instead of the backpropagation (BP) learning adopted by many existing methods. The consequent parameters are obtained using the. least squares estimates algorithm. The NF-HLA is employed in a static function approximation and in nonlinear system identification. Simulation results demonstrate that a compact and high-performance neuro-fuzzy system can be constructed. Comprehensive comparisons with other approaches show that the proposed approach is superior in terms of learning efficiency and performance.