Artificial neural networks design based on modified adaptive particle swarm optimization
Jui-Chuan Cheng, Te‐Jen Su, Ming-Yuan Huang, Chi-Yuan Juang · 2011
This paper presents a weights training method of the artificial neural networks (ANN), which combines modified adaptive particle swarm optimization (MAPSO) with Back-propagation (BP) to apply to function approximation. BP is an approximate steepest descent algorithm, hence some inherent problems are frequently encountered in the use of this algorithm, e.g., very slow convergence speed in training, easily to get stuck in a local minimum, etc. This study uses the particle swarm optimization (PSO) method to avoid this problem. From the demonstrated examples, compared with PSO-ANN, APSO-ANN, we have obtained the better performance, better approximation and less convergence generations from the proposed MAPSO-ANN.