Stefan Steinerberger, University of Washington

Bad Science Matrices
Date
Oct 21, 2024, 12:25 pm1:25 pm

Details

Event Description

We will consider the problem of a bad scientist trying to design a battery of (individually fair) statistical tests such that a fair unbiased coin is typically identified as biased by at least one test.  Geometrically, this corresponds to finding n x n square matrices that send the 2^n vertices of the {-1,1}^n cube to points with at least one large coordinate on average. We discuss an asymptotic solution of the problem and the curious structure of low-dimensional extremal examples when n=2,3,4,.... Connections to vector-balancing problems and the Komlos conjecture are discussed.

Short Bio: Stefan Steinerberger is a Professor of Mathematics at the University of Washington, Seattle. He is interested in analysis and its applications in many different settings and has a weakness for simple-sounding problems. His research has been supported by INET, NSF and the Alfred P. Sloan Foundation.

Event Category
S. S. Wilks Memorial Seminar in Statistics