Particle swarms for feedforward neural network training

Particle swarm is an optimization paradigm for real-valued functions, based on the social dynamics of group interaction. We propose its application to the training of neural networks. Comparative tests were carried out, for classification and regression tasks.