A hybrid approach for parameter optimization of RBF-AR model

A hybrid global-local optimization algorithm for radial basis function (RBF) networks and RBF nets-based state-dependent autoregressive (RBF-AR) models parameter estimation is presented. This algorithm (EA-SNPOM) effectively combines an evolutionary algorithm (EA) with a gradient-based search strategy named the structured nonlinear parameter optimization method (SNPOM). The hybrid approach provides a global search with the EA and a local search via the SNPOM. The effectiveness of the resulting combination is demonstrated by several examples.