Ridge regression learning in ESN for chaotic time series prediction
Min Han · Kongzhi yu juece · 2007
As a new type of recurrent neural network,echo state network(ESN) is applied to nonlinear system identification and chaotic time series prediction.A technique is proposed to improve the properties of ESN solution,which performs ridge regression in the reservoir state space instead of the previous linear regression.In addition,the ridge regression parameter is systematically determined by Bayesian method or Bootstrap method.The prediction results for monthly sunspots time series show a satisfying performance.