Olivetti Club

Diana HalikiasCornell University
Matrix Recovery and Continuous Analogues

Tuesday, November 8, 2022 - 5:15pm
Malott 406

The importance of recovery is likely the first lesson you would learn from top-rope rock climbing. I will introduce you to the idea of recovery in a different context: that of learning black-box matrices with unknown entries, which we can only interact with via matrix-vector multiplication. We will explore randomized, numerical, and exact algorithms to recover various structured matrices, as well as results that describe the complexity of these structures. Next, since hierarchical matrices are a particular class of matrices that approximate Green’s functions, we will spend some time on the continuous analogue of this problem, ending with a short discussion of the theory of elliptic PDE learning.