1 h
Seminar, Hosted by Simons Foundation
Randomized Methods for Very Large-Scale Linear Algebra
Professor of Mathematics, University of Texas at Austin
Recording
Abstract
At the core of many computational simulations of physical systems lies a linear algebraic problem, such as a system of linear equations or an eigenvalue problem, and similar problems arise throughout computational statistics and data science. Because these computations are often the most expensive part of a simulation, they frequently determine both its feasibility and the level of physical detail it can achieve.
This Presidential Lecture describes how ideas from random matrix theory and high-dimensional probability have enabled a new generation of algorithms that are far faster than traditional deterministic methods and make previously intractable simulations possible. A key feature of these randomized methods is that they require very little data movement, which suits them to modern computing hardware. The lecture presents methods for low-rank approximation, for solving linear systems, and for constructing highly accurate compressed representations of continuum operators, including solution operators for elliptic partial differential equations, Dirichlet-to-Neumann maps and time-evolution operators.