The echo state approach on nonlinear system modelling
"Echo State Networks" (ESNs) is a new approach of training Recurrent Neuronal Networks. ESNs enable the use of large nets having several thousands internal nodes. These large nets have the capacity of learning high-dimensional non-linear temporal patterns. In this paper, the use of ESNs is demonstrated on learning the input output behavior of three benchmark systems provided by the organizers of the DoE conference.