Seminar | March 16 | 5:15-6:15 p.m. | 891 Evans Hall
Bernd Sturmfels, UC Berkeley and MPI Leipzig
Mixtures of Gaussians are ubiquitous in data science. We give an introduction to the geometry of these statistical models, with focus on the tensors that represent their higher moments. The familiar theory of rank and border-rank for symmetric tensors is recovered when all covariance matrices are zero. Recent work with Carlos Amendola and Kristian Ranestad characterizes the circumstances under which Gaussian mixtures are identifiable from their moments.