Nonlinear dynamic system identification with dynamic recurrent neural networks
Gustavo Calderón, Jean-Philippe Draye, Davor Pavisic, Roberto Teran, G. Libert · 2002
We work with the dynamical recurrent neural network as a tool for system identification. We train this network using a time-dependent back-propagation learning algorithm and we show that for modeling a nonlinear dynamical system, our neural device has good performance for interpolation and extrapolation, and is very robust in the presence of noise.