Kristian Ranestad (Oslo): Gaussian mixtures

Abstract: The Gaussian moment variety is a noisy version of the Veronese variety, and parametrises higher order moments given a mean vector and a covariance matrix for a multivariate Gaussian distribution.  Gaussian mixtures are linear combinations of higher order moments, i.e. points on higher order secant varieties of the moment variety.  I will report on recent work with Carlos Amendola and Bernd Sturmfels on the dimension of these secant varieties, in particular whether a general mixture is the mixture of a unique set of moments..
Published Mar. 20, 2017 10:05 PM - Last modified Mar. 20, 2017 10:05 PM