Adjusting Weights and Architecture of Neural Networks through PSO with Time-Varying Parameters and Early Stopping

This paper presents results of an approach to optimize architecture and weights of MLP Neural Networks, which is based on particle swarm optimization with time-varying parameters and early stopping criteria. This approach was shown to achieve a good generalization control, as well as similar or better results than other techniques, but with a lower computational cost, with the ability to generate small networks and with the advantage of the automated architecture selection, which simplify the training process.