Applied Analysis Seminar

Tuesday July 23 in K9509, SFU

Speaker and Title: 2:00 - 2:45: W. Hare (SFU) "Identifing Manifolds for Partially Smooth Functions"

Abstract: Given a weak set of demands we show that manifolds associated with Partially Smooth functions can be identified in a finite number of steps.

Speaker and Title: 2:45 - 3:30: Adrian Lewis (SFU) "Robust regularization in optimization"

Abstract: Suppose we want to minimize a function that is evaluated not at the point we choose, but rather at an unpredictably perturbed point. In other words, we want to minimize a ``robust regularization'': at any point, this regularization equals the maximum value of the function in a fixed neighbourhood of the point. For example, a crucial function in feedback control is the spectral abscissa of a square matrix (the largest real part of an eigenvalue). Its robust regularization is the ``pseudospectral absicssa'', a useful tool in robust control. I will describe conditions guaranteeing that the robust regularization is locally Lipschitz even when the underlying function is not, and apply this result to the pseudospectral abscissa.