Assistant Professor, University of Wisconsin–Madison
Title
Spectral Methods in Contemporary Multivariate Analysis
Abstract
Spectral methods, namely techniques based on the study of matrix eigen-properties, factorizations, and perturbations, have experienced a renaissance over the past fifteen years, fueled by sustained activity in multiple disciplines including Statistics, Probability, and Computer Science. This talk briefly highlights the current and foreseeable future role of spectral methodology in modern-day multivariate analysis. Several recent results will also be discussed in detail, pertaining to graph clustering, matrix de-noising, and sparse latent factor estimation.
Bio
