Volterra prediction models and higher order whiteness

For nonGaussian processes, a nonlinear predictor can achieve a smaller prediction error than a linear one. The authors study Volterra predictors and compare them with their linear counterparts. The concept of Volterra unpredictability leads to important generalizations of the notion of white noise to higher orders. These generalizations are introduced and relations are established between them. Examples are provided to show that the relations between these higher order whitenesses are not trivial.<<ETX>>