Natural gradient learning neural networks for modeling and identification of nonlinear systems with memory

This paper applies natural gradient (NG) learning neural networks (NNs) for modeling and identification of nonlinear systems with memory. The nonlinear system is comprised of a discrete-time linear filter H followed by a zero-memory nonlinearity g(.). The neural network model is composed of a linear adaptive filter Q and a two-layer nonlinear neural network (NN). It is shown that the NG learning method outperforms the ordinary gradient descent method in terms of convergence speed and mean squared error (MSE) performance.