Nonparametric identification of quasi-stationary systems
Authors:
Leszek Rutkowski
Affiliation:
Department of Electrical Engineering, Technical University of Cz?stochowa, Zawadzkiego 17, 42-200 Cz?stochowa, Poland
Abstract:
Recursive estimates of a regression function — based on the Parzen-Rosenblatt kernels — are applied for identification of quasi-stationary systems. It is shown that these estimates do not lose their asymptotic properties even if the data are not stationary. Conditions for weak and strong pointwise consistency are given.