Review on echo state networks

The echo state network(ESN) is a novel kind of recurrent neural network and has recently become a hot topic for its easy and distinctive training method along with high performance.In ESN,the reservoir computing method is introduced,which is a completely new approach used to design a recurrent neural network.By comparing this novel model with existing recurrent neural network models,it can overcome the difficulty encountered in slow convergence and local minimum in the gradient descent based training algorithm.Currently,there is considerable enthusiasm for the research and application of ESN.A review on ESN is presented in this paper.The advantages and drawbacks of ESN and various improvements are analyzed.Finally,some future research directions are also discussed.