Echo State Network prediction based on Backtracking Search optimization Algorithm

Abstract

Echo state network is one of the latest research results of artificial neural network, it’s most notable feature is the simple network structure. Only the connection weight matrix of the reservoir (hidden layer) to the output layer needs to be calculated. In this paper, we aim to provide a kind of the echo state network prediction method based on adaptive backtracking search optimization. The adaptive back search algorithm is introduced into the echo state network to optimize the connection weights matrix of the echo state network reservoir and the output layer, then to improve the echo state network prediction accuracy.