Oberwolfach Reports

Full-Text PDF (538 KB) | Introduction as PDF | Metadata | Table of Contents | OWR summary
Volume 10, Issue 1, 2013, pp. 239–293
DOI: 10.4171/OWR/2013/04

Published online: 2013-11-02

Numerical Methods for PDE Constrained Optimization with Uncertain Data

Matthias Heinkenschloss[1] and Volker Schulz[2]

(1) Rice University, Houston, United States
(2) Universit├Ąt Trier, Germany

Optimization problems governed by partial differential equations (PDEs) arise in many applications in the form of optimal control, optimal design, or parameter identification problems. In most applications, parameters in the governing PDEs are not deterministic, but rather have to be modeled as random variables or, more generally, as random fields. It is crucial to capture and quantify the uncertainty in such problems rather than to simply replace the uncertain coefficients with their mean values. However, treating the uncertainty adequately and in a computationally tractable manner poses many mathematical challenges. The numerical solution of optimization problems governed by stochastic PDEs builds on mathematical subareas, which so far have been largely investigated in separate communities: Stochastic Programming, Numerical Solution of Stochastic PDEs, and PDE Constrained Optimization. The workshop achieved an impulse towards cross-fertilization of those disciplines which also was the subject of several scientific discussions. It is to be expected that future exchange of ideas between these areas will give rise to new insights and powerful new numerical methods.

No keywords available for this article.

Heinkenschloss Matthias, Schulz Volker: Numerical Methods for PDE Constrained Optimization with Uncertain Data. Oberwolfach Rep. 10 (2013), 239-293. doi: 10.4171/OWR/2013/04