Abstract:A novel training algorithm for RBF(Radical Basis Function) networks was proposed based on extended Kalman filter and fuzzy control.The simulation results were presented on RBF networks as applied to the Iris classification problem.It is shown that the use of the extended Kalman filter and fuzzy control results in faster learning and higher precision of identification than conventional RBF networks.

A Novel Learning Algorithm for RBF Network Based on Kalman Filter and Fuzzy Control