Thurs. May 8 at 3:00pm in FRET 402

Mikhail Malioutov,  Department of Mathematics, Northeastern University

Abstract:  Modeling applications such as literary texts as stationary time series is a firmly established tool in computer linguistics. A popular problem is an authorship attribution of texts which in statistical terms is called homogeneity testing which must be nonparametric with respect to the multivariate distribution of time series since data will never suffice to find their parametric model. An adequate tools turn out to be Variable memory Length MC and Conditional Complexity of  Compression. The latter was initially inspired by analogy with non-computable Kolmogorov Complexity. A review of the underlying theory and many applications will be presented.