Wavelet frames based estimator

Abstract

This paper introduces a new wavelet frames-based functional estimation method (i.e. a wavelet-based neural network) which works for more than one dimension functions. The use of frames and wavelets in our approach yields to robust decomposition with an interesting parsimonious property: compression of information in few coefficients. This approach is illustrated using the problem of estimating radioactivity in Chernobyl area.