Rachel Ward, University of Texas
We present new spectral-norm concentration inequalities for certain normalized products of random matrices. which arise in the analysis of stochastic iterative algorithms such as Oja's algorithm for streaming Principal Component Analysis (PCA). Besides being of independent mathematical interest, our results provide a new, direct proof of convergence for Oja's algorithm.
Rachel Ward is the W.A. "Tex" Moncrief Distinguished Professor in Computational Engineering and Sciences — Data Science and Associate Professor of Mathematics at UT Austin. From 2017-2018, she was a visiting research scientist at Facebook AI Research. She is recognized for her contributions to sparse approximation, stochastic optimization, and numerical linear algebra. Prior to joining UT Austin in 2011, Dr. Ward received the PhD in Computational and Applied Mathematics at Princeton in 2009 and was a Courant Instructor at the Courant Institute, NYU, from 2009-2011. Among her awards are the Sloan research fellowship, NSF CAREER award, and the 2016 IMA prize in mathematics and its applications.