The power of randomness in scientific computing
Thursday, February 27, 2020
3:30PM – 5PM
Most numerical methods in scientific computing are deterministic. Traditionally, accuracy has been the target while the cost was not the concern. However, in this era of big data, we incline to relax the strict requirements on the accuracy to reduce numerical cost. Introducing randomness in the numerical solvers could potentially speed up the computation significantly at small sacrifice of accuracy. In this talk, I'd like to show two concrete examples how this is done: first on random sketching in experimental design, and the second on numerical homogenization, hoping the discussion can shed light on potential other applications. Joint work with Ke Chen, Jianfeng Lu, Kit Newton and Stephen Wright.
Dr. Qin Li is an Associate Professor of Mathematics at the UW-Madison and also holds an affiliation with The Wisconsin Institutes for Discovery. Her general research interests lie in numerical analysis and scientific computing. https://www.math.wisc.edu/~qinli/