Schwerpunktskolloquium Analysis und Numerik: All-At-Once versus Reduced Iterative Methods for Time Dependent Inverse Problems

Wann
Montag, 4. Dezember 2017
17 bis 18 Uhr

Wo
F 424

Veranstaltet von

Vortragende Person/Vortragende Personen:
Professor Dr. Barbara Kaltenbacher (Alpen Adria Universität Klagenfurt, Österreich)

Diese Veranstaltung ist Teil der Veranstaltungsreihe „Schwerpunktskolloquium Analysis und Numerik“.

Abstract: A large number of inverse problems in applications ranging from engineering via economics to systems biology can be formulated as a state space system with a nite or in nite dimensional parameter that is supposed to be identi ed from additional continuous or discrete observations. The aim of this talk is to formulate inverse problems of this kind in a reduced and an all-atonce fashion and compare some iterative regularization methods in these two di erent settings as they have already been shown to exhibit crucial di erences for stationary inverse problems. In particular, we will consider Landweber iteration and the iteratively regularized Gauss-Newton method, as well as the Landweber-Kaczmarz iteration, as the latter obviously lends itself to a subdivision of the problem to a union of time intervals.