Research on design method of small world property ESN

Authors:
Yingchun Bo;Junfei Qiao;Shuwei Wang
Affiliations:
College of Electronic and Control Engineering, Beijing University of Technology, Beijing, China;College of Electronic and Control Engineering, Beijing University of Technology, Beijing, China;College of Architecture and Civil Engineering, Beijing University of Technology, Beijing, China
Venue:
ISNN'11 Proceedings of the 8th international conference on Advances in neural networks - Volume Part I
Year:
2011

Citing 4
Cited 0

Analysis and design of echo state networks

Neural Computation
2007 Special Issue: Decoupled echo state networks with lateral inhibition

Neural Networks
Stability of small-world networks of neural populations

Neurocomputing
Large-scale neural systems for vision and cognition

IJCNN'09 Proceedings of the 2009 international joint conference on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

Aim to solve the problems of structure design and parameters selection about conventional ESN, a small world property ESN (SWESN) is proposed in this paper. Neuron space growth algorithm is adopted to generate a physical network with small world topology on 2-D plane firstly, and then the nodes and the connections of the physical network are mapped into the neurons in reservoir of SWESN, Thus the dynamic neuron reservoir (DNR) in SWESN has small world characteristic. In addition, different typical neurons are adopted in the reservoir. The simulation experiments confirms that the SWESN generated by this method could create more abundant dynamic behavior than conventional ESN, and SWESN exceeds conventional ESN both at robustness and at anti-disturbance ability.