A nonlinear recurrent neural network is trained to synthesize chaotic signals. The identification process is reduced to a teaching phase and a linear regression. The influence of the shape of the nonlinearity in the neurons and the noise amplitude are stud ...
2003