On optimality conditions for nonsmooth functions
8. November 2023 um 17:15 Uhr
Ort:
Uhrturmhörsaal (S2|08 171), Hochschulstraße 4.
Vor dem wissenschaftlichen Vortrag trifft sich die Vortragende mit Promovierenden und Postdocs am Anfang ihrer Karriere und teilt persönliche Erfahrungen aus ihrem Werdegang.
Abstract:
Numerous optimization tasks exhibit a nonsmooth behavior. In contrast to the classical smooth case, where optimality conditions are well studied and understood, criteria to determine whether a given point is optimal or even just stationary are still the subject of ongoing research for nonsmooth functions to be minimized.
In this presentation, first we present already established optimality conditions based on generalized derivative concepts for unconstrained nonsmooth problems. Subsequently, we discuss new optimality conditions for a large class of piecewise smooth functions using so-called kink qualifications. Here, also the computational complexity to verify the new criteria is covered. Finally, we show the connections of the kink qualifications to the constraint qualifications for MPECs.
Bio:
Andrea Walther studied business mathematics at Universität Bayreuth and got her PhD from TU Dresden in 1999, from where she also received her habilitation in 2008. In 2009, she became professor for Mathematics and its applications at Universität Paderborn. Since October 2019, Andrea Walther is MATH+ Professor for Mathematical Optimization at the Humboldt-Universität zu Berlin.
Her research interests are in the field of nonlinear optimization with a focus on adjoint-based optimization and non-smooth optimization as well as in algorithmic differentiation.
She is member of the board of the International Association of Applied Mathematics and Mechanics (GAMM).
