Wavelet reconstruction of nonlinear dynamics

We investigate the reconstruction of embedded time-series from chaotic dyna mical systems using wavelets. The standard wavelet transforms are not appli cable because of the embedding, and we use a basis pursuit method which on its own does not perform very well. When this is combined with a continuous optimizer, however, we obtain very good models. We discuss the success of this method and apply it to some data from a vibrating string experiment.