Neural-based identification for nonlinear dynamic systems

In this paper, a time-delayed feed-forward neural network is used to make a time-domain characterization of the nonlinear dynamic behavior of an electronic device. The procedure provides also an analytical expression as a Volterra Series model. This model, however, can be built to different accuracy degrees, depending on the activation function chosen for the neural network used. Three Volterra series models extracted from different neural networks are compared, having hyperbolic tangent, cubic and quadratic activation functions. This analysis is applied to the modeling of a nonlinear Power Amplifier (PA).