Newton-basierte Methoden für restringierte Optimierungsprobleme mit partiellen Differentialgleichungen

Date

Jan 15, 2014

Time

5:00 PM - 6:00 PM

Speaker

Prof. Dr. Michael Ulbrich

Affiliation

TU München, Lehrstuhl für Mathematische Optimierung

Series

TUD Dresdner Mathematisches Seminar

Language

de

Main Topic

Mathematik

Other Topics

Mathematik

Description

The last ten years have seen significant advances in the development and analy-sis of efficient methods for solving inequality constrained optimization problems governed by partial differential equations (PDEs). In particular, semismooth Newton methods are now established as a standard approach for PDE constrained optimization. The availability of a convergence theory in function space is a particular strength that makes algorithms especially well suited for PDE-constrained problems since this forms the basis for mesh independent convergence behavior. In this talk, we focus on two important classes of Newton-based methods for PDE-constrained optimization: Semismooth Newton and inte-rior point methods. We discuss the particular challenges in comparison to finite dimen-sional settings and present selected aspects about the development, convergence analy-sis, and properties of these methods.

Links

• Dresdner Mathematisches Seminar

Last modified: Jan 9, 2014, 3:06:08 PM