Improving learning enhancement radial basis function neural network using improved harmony search algoritim

Abstract

Radial Basis Function (RBF) neural network training with Particle Swarm Optimization (PSO) overcomes the trapping to the local minimum by Back Propagation (BP) algorithm and slow computation of Genetic Algorithm (GA). However, PSO converged too fast which makes it to be trapped in the local optimum. Furthermore, particles may move to an invisible region. Therefore, to enhance the learning process of RBF and overcome the problem associated with PSO, Improved Harmony Search Algorithm (IHSA) was employed to optimize the RBF network to enhance its learning capacity. The study conducted performs comparative analysis between the hybrid of IHSA and RBF network and the PSO-RBF network. The results obtained show that IHSA has increased the learning capability of RBF neural network in terms of correct classification percentage and error convergence rate. The proposed IHSA-RBF model gives higher performance with promising results compared to PSO-RBF network