Nonlinearity recovery by standard and aggregative orthogonal series algorithms

In this paper, the problem of nonlinearity recovery in Hammerstein systems is considered. Two algorithms are presented: the first is a standard orthogonal series algorithm, whereas the other, ie, the aggregative one, exploits the convex programming approach. The finite sample size properties of both approaches are examined, compared, and illustrated in a numerical experiment. The aggregative algorithm performs better when the number of measurements is comparable to the number of parameters; however, it also imposes additional smoothness restrictions on the recovered nonlinearities.