PSO as an effective learning algorithm for neural network applications

This paper introduces an improved particle swarm optimization (PSO) as a new tool for training an artificial neural network (ANN). As a consequence, an accurate comparison with other optimization methods is needed; the typical supervised feed-forward backpropagation algorithm (EBP) and the classical genetic algorithm (GA) are chosen. The aim is to highlight advantages and drawbacks of PSO technique in order to suitably apply it to neural network applications in electromagnetic problems. Some numerical results and comparisons are presented analyzing a load forecasting problem.