A hybrid learning algorithm combined with generalized RLS approach for radial basis function neural networks

摘要

In this paper, a new hybrid learning method for radial basis function neural networks based on generalized recursive least square algorithm is proposed. Firstly the generalized recursive least square (GRLS) model including a general quadratic weight decay term in the energy function for the training of RBF neural networks is described. Then combined with the GRLS approach, a new hybrid learning method is proposed to meet the design goals: improving the generalization ability of the trained network. Finally experimental results demonstrate that our approach can achieve a significantly improved generalization performance of the RBF networks.