Approach for optimizing echo state network training based on PSO

Echo state network(ESN) is a new architecture and learning method of recurrent neural network.Compared to traditional neural network it has much stronger ability of non-linear prediction,however it also can give rise to more samples required to be trained.The proposed model called PSO-ESN used the PSO algorithm to train the output connect weight and the prediction precision is effectively increased with limit samples.The algorithm also is verified by the experiment on prediction of network traffic data.