Numerical Algebraic Geometry in Algebraic StatisticsAlgebra & Discrete Mathematics
|Jose Rodriguez, UC Berkeley, Dept. of Mathematics
|Thu, Mar 21 2013, 3:10PM
Maximum likelihood estimation is a fundamental computational task in statistics and involves beautiful geometry. We discuss this task for determinantal varieties (matrices with rank constraints) and show how numerical algebraic geometry can be used to maximize the likelihood function. Our computational results with the software Bertini led to surprising conjectures and duality theorems. This is joint work with Jan Draisma, Jon Hauenstein, and Bernd Sturmfels.