Joshua Agterberg – “Nonidentifiability and nonparametric random graph hypothesis testing”
Registration required at: https://wse.zoom.us/meeting/register/uJIof-ytqz8ihv8nenBvEQBO_kUcmxyUUA
Abstract: Hypothesis testing for random graphs is a relatively new field, and existing methods have focused mostly on specific random graph models, such as the stochastic block model and its variants. However, there are not currently many consistent hypothesis tests for general low-rank random graphs, which suffer from a unique form of nonidentifiability. In this talk I will be discussing two types of nonidentifiability that arise when using spectral methods to study random graphs, and I will show how limiting results for one type of nonidentifiability can be used to find a consistent nonparametric hypothesis test for equality of distribution for general low-rank random graphs.
Bio: Joshua Agterberg is a PhD student in Applied Mathematics and Statistics at Johns Hopkins University where he is advised by Carey Priebe. His research interests include statistical inference for random graphs, kernel methods, spectral perturbation theory, high-dimensional statistics, and nonparametric statistics, with an emphasis on the interplay between optimization, statistics, and matrix analysis. This year, he is a MINDS fellow, a Counselman fellow, and an AMS apprentice teaching fellow.