Cuckoo Search Based Optimization of Echo State Network for Time Series Prediction

Abstract

The Echo State Network (ESN) is a type of reservoir computing (RC) approach of artificial neural networks. The choice of its parameters and topology have become an optimization problem. This paper proposes a cuckoo search based optimization of three global parameters of the echo state network for time series prediction. The proposed technique is compared with the classical echo state network on the prediction of Mackey-Glass time series. It was found that the cuckoo search based method outperformed the classical ESN.