The Reservoir Topology of Echo State Network

Echo state network (ESN) has a unique reservoir topology and is widely used to solve a variety of problems. However, the relationship between the reservoir topology and the performance of the network is unclear. To address this problem, based on the small-world (SW) network and modularity theory, the small-world echo state network (SWESN) and modularity ESN are used to explore the relationship between the reservoir topology and the prediction performance of the network by experimental analysis. The experimental results show that the SW topology and random modularity topology of a reservoir play an important role in improving the prediction performance of the network. Moreover, the modularity echo state network with random topology (M-ESN) shows higher prediction accuracy.