A0834
Title: Nonparametric maximum likelihood and related methods in infinite-dimensional situations: Convex optimization aspects
Authors: Ivan Mizera - University of Alberta (Canada) [presenting]
Abstract: Certain nonparametric methods, including, but not limited to nonparametric maximum likelihood, and applied to problems like density estimation in shape-constrained situations, nonparametric estimation of mixture models, and others, are reviewed from the theoretical point of view. These methods have been already demonstrated to work in practical problems; the focus of the present analysis, rather than statistical properties - which have been to some extent analyzed in the literature elsewhere - is on the aspects of convex optimization employed in their implementation, in particular on the theoretical vindication of certain approximate strategies, possible aspects of regularization, and the potential of extensions to high-dimensional situations. Some practical details are discussed as well and illustrated on data-analytic examples.
