Free Information Geometry

This project will develop an information geometry for free probability theory.

Free probability is a theory of non-commuting random variables that describes the large-n behavior of many families of n x n random matrices.

Free probability has had applications to data analysis, communication, finance, and many other topics where matrices appear, as well as to the structure of von Neumann algebras, which is a wide-ranging and challenging field of pure mathematics.

Information geometry refers to the synthesis of optimal transport together with measures of information such as entropy, which has shaped much recent work in partial differential equations, optimization, and data analysis.

Information geometry has motivated many corresponding results in free probability theory, but several deep questions remain open concerning the relationship between optimal transport and entropy in free probability, and whether it accurately describes the large-n limit of the classical information geometry for random matrices.

This project aims to show that free entropy is concave along optimal transport geodesics, establish the existence of momentum measures in free probability, give an optimal control formulation of free entropy, and exhibit counterexamples to regularity properties for optimal transport through connections with quantum information.

The project will be supervised by Magdalena Musat in the Department of Mathematical Sciences at the University of Copenhagen, which provides a wealth of training and resources on operator algebras and quantum mathematics.