Harmony Search Algorithm with Chaos for Training RBFNN

Harmony search algorithm is a heuristic optimization method inspired from the improvisation process of musicians. A new version of harmony search algorithm with chaos is proposed. This algorithm initializes and updates the harmony memory with the chaos optimization algorithm based on secondary carrier wave, adjusts the parameters dynamically to improve the convergence rate and optimization accuracy. In this paper this new algorithm is used to train RBF neural network. The training performance and generalization capability of the new algorithm are tested and verified on function approximation and Iris classification problem. The training time, MSE and accuracies of the new algorithm are compared with standard HS, IHS, GHS algorithms. It turns out that the new algorithm has better convergence rate and accuracy than others. Finally, we applied the new algorithm into the sewage treatment water quality prediction. The relative errors between the actual values and predictive values prove that a RBFNN trained by the new algorithm has better performance than others.