Abstract
We study data-fitting methods based on wavelet shrinkage. We discuss several methods for threshold and non-treshold shrinkage of the wavelet coefficients, and compare their performance for the model problems of non-parametric regression and density estimation, under the assumption of Besov regularity of the estimated function. The results are illustrated by graphical visualization of several numerical examples, using the new wavelet-based modelling and visualization library GM-waves.
