On Tikhonov regularization, bias and variance in nonlinear system identification (Q1356153)

For nonlinear model identification, the author uses penalty terms such as in the Tikhonov theory, for solving ill-posed or ill-conditioned identification problems. In this work, it is proved that the Tikhonov regularization method is a powerful alternative for regularization of nonlinear system identification problems by introducing smoothness of the model as a prior. Some properties of this regularization techniques are studied and discussed in terms of an analysis of bias and variance, and illustrated by a semirealistic simulation example.