TRAINING PARTIALLY RECURRENT NEURAL NETWORKS USING EVOLUTIONARY STRATEGIES

This correspondence presents the latest results of using evolutionary strategies (ES's) to design partially recurrent neural networks for vi seme recognition. ES's are stochastic optimization algorithms based up on the principles of natural selection found in the biological world. Our results indicate that ES's can be effectively used to determine th e synaptic weights in neural networks and can outperform backpropagati on techniques.