Data clustering and noise undressing of correlation matrices

M. Marsili

We discuss a new approach to data clustering. We find that maximum likelyhood leads naturally to an Hamiltonian of Potts variables which depends on the correlation matrix and whose…