Local correlation's potential for noise reduction and symbolic partitions

We investigate local correlation dimension-based noise-cleaning of time ser ies, where points having anomalously large dimensions are iteratively remov ed from the reconstructed attractor. We find an optimal rang for the number of iterations in which the algorithm yields good results. Choosing non-loc al ranges for the linear regression yields a new method for finding nonhype rbolic tangency points. The method is also applicable for noisy systems wit h unknown dynamics; in this case, noise facilitates the detection of the po ints.