Costruzione di Stimatori per dati in spazi multidimensionali

15 Luglio 2011

This research project is intended to continue an ongoing collaboration that has had a strong record of publications in the past few years and for which we see several directions for further developments. This project proposes viable alternatives to the existing methods for estimating filaments and manifold in high dimensional spaces. The procedures available now are characterized by strong theoretical properties, but they are not operationally accessible. We plan to develop 1. New methods for estimating hidden features in massive datasets, theoretically justified and of simple implementation. 2. New computational techniques to allow wide use of these estimators in diverse applied disciplines. 3. New statistical theory for exploring the features of the new estimators proposed.