Applied Algebra and Geometry Seminar: Janike Oldekop
Janike Oldekop from TU Berlin will give the following talk:
Title: Likelihood Geometry of Toric Models
Abstract: Toric models form an important subclass of discrete statistical models that are naturally described by toric varieties, making them interesting objects of study in algebraic statistics. In this talk, we investigate the likelihood geometry of these models, addressing the fundamental question of how geometric and combinatorial properties of a toric model influence maximum likelihood (ML) estimation. Our main focus is on the ML degree, an algebraic complexity measure of the underlying optimization problem. We survey recent developments and highlight current research directions, which connect algebraic ML estimation with the theory of A-discriminants introduced by Gelfand, Kapranov, and Zelevinsky.