Numerical Algebraic Geometry for Maximum Likelihood Estimation

Numerical algebraic geometry is a growing area of applied algebra that involves describing solutions of polynomial systems of equations. This area has already had an impact in kinematics, statistics, PDE's, and pure math. In this talk we focus on using numerical algebraic geometry for maximum likelihood estimation in algebraic statistics.

In the first half, we will introduce a concrete example and develop background information. In the second half, we present computational results consisting of Maximum Likelihood degrees for matrices with rank constraints and present the surprising theoretical result of maximum likelihood duality.

Event Topic

Nonlinear Algebra and Statistics (NLASTATS)