Diplomanden- und Doktorandenseminar
Numerische Mathematik SS 2006


Dozenten: Jörg Liesen, Christian Mehl, Volker Mehrmann, Reinhard Nabben, Harry Yserentant
Koordination: Christian Mehl
LV-Termine:Do 10:15-11:45 in MA 313
Inhalt: Vorträge von Diplomanden, Doktoranden, Postdocs und manchmal auch Gästen zu aktuellen Forschungsthemen

Vollständige Terminplanung:
 
Datum Uhrzeit Raum Vortragende(r) Titel
Do 21.4.2006 10:15 MA 376
 
  Vorbesprechung
Do .4.5.2006 10:15 MA 376
 
Simone Bächle Minimal extension for DAEs from charge and flux oriented MNA
Do 11.5.2006 10:15 MA 376
 
Martin Bodestedt The implicit function theorem and perturbation analysis of differential-algebraic equations
Do 18.5.2006 10:15 MA 376
 
Marion Rauscher Besov Spaces Part I (Abstract)
Do 18.5.2006 im An-
schluss
MA 376
 
Hans-Christian Kreusler Besov Spaces Part II (Abstract)
Do 1.6.2006 10:15 MA 376
 
Christian Schröder A QR-like algorithm for the palindromic eigenvalue problem (Abstract)
Do 1.6.2006 im An-
schluss
MA 376
 
Elena Virnik An SVD approach to identifying meta-stable states of Markov chains (Abstract)
Do 8.6.2006 10:15 MA 376
 
Sadegh Jokar Sparsity and Beyond (Abstract)
Do 15.6.2006 10:15 MA 376
 
Lena Wunderlich Structure preserving condensed forms for pairs of Hermitian matrices and matrix valued functions (Abstract)
Do 22.6.2006 10:15 MA 376
 
Andreas Zeiser Introduction to wavelets and their applications (Abstract)
Do 6.7.2006 10:15 MA 376
 
Christian Mense Multilevel Methods for singular Matrices (Abstract)
Do 6.7.2006 im An-
schluss
MA 376
 
Lisa Poppe Delay Differential Algebraic Equations (Abstract)

Interessenten sind herzlich eingeladen!


Weitere Vorträge siehe auch:

Rückblick:


Abstracts zu den Vorträgen:


Marion Rauscher und Hans-Christian Kreusler (TU Berlin)
Besow Spaces Part I and Part II
Do 18.5.2006, 10:15 Uhr in MA 376

Abstract:
There are different ways to define Besov spaces. In the first part the classical method using moduli of smoothness is presented whereas in the second part Besov spaces are introduced in the more general framework of interpolation theory. However both definitions lead to identical spaces as the corresponding norms are equivalent.


Christian Schröder (TU Berlin)
A QR-like algorithm for the palindromic eigenvalue problem
Do 1.6.2006, 10:15 Uhr in MA 376

Abstract:
The talk is on an adapted version of the QR algorithm to transform a square matrix to 'lower-right triangular' form by unitary congruence. The talk is in English. The foils are in German. Interactive matlab examples are planned.


Sadegh Jokar and Marc Pfetsch (TU Berlin)
Sparsity and Beyond
Do 8.6.2006, 10:15 Uhr in MA 376

Abstract:
The signal processing literature has been interested in finding sparse solutions of underdetermined linear equation systems. This can be formulated as the minimization of the l_0 quasi-norm. Several researchers have investigated the approximation of this optimization problem with l_1 minimization. In this talk, first we will review some of the recent results in this field. Then we will empirically investigate the exact computation of l_0 minimization problems via branch-and-cut based on reduction to the so-called Maximum feasible subsystem problem. We will also present several heuristics like an adaptation of an algorithm by Mangasarian. With our exact method, we are able for the first time to evaluate the performance of the l_1 minimization and other heuristics for computing approximate l_0 minimal solutions. In particular, we will investigate the behavior of the l_1 minimization when the sparsity of the solutions lies beyond the theoretical limits for which an exact recovery is guaranteed.


Lena Wunderlich (TU Berlin)
Structure preserving condensed forms for pairs of Hermitian matrices and matrix valued functions
Do 15.6.2006, 10:15 Uhr in MA 376

Abstract:
The study of matrix pairs or pairs of matrix valued functions is often motivated by applications from linear differential-algebraic equations. In many applications from mechanics or control theory the underlying matrices are symmetric or Hermitian. We present a condensed form under structure preserving congruence transformations for pairs of Hermitian matrices. Further, we show that for pairs of Hermitian matrix valued functions a condensed form only exists if certain additional assumptions hold. There only exists a structure preserving normal form if the corresponding differential-algebraic system is of index less than or equal to 1.


Andreas Zeiser (TU Berlin)
Introduction to wavelets and their applications
Do 22.6.2006, 10:15 Uhr in MA 376

Abstract:
The intention of this talk is to introduce the basic ideas of wavelets. The theory of wavelets will be discussed and application like signal processing and image compression will be briefly explained. However, the main focus will lie on the application of wavelets to the discretization of PDEs.

Literature:


Christian Mense (TU Berlin)
Multilevel Methods for singular Matrices
Do 6.7.2006, 10:15 Uhr in MA 376

Abstract:
In this talk we are discussing solution methods for singular systems Ax=b with b in the range of A. One application, where these singular systems arise in, is the simulation of the human blood flow as a Markov chain. Beside easy to satisfy conditions to guarantee convergence for general stationary iteration schemes, we are presenting convergence results for the AMLI- and the MAMLI-methods.


Impressum Christian Mehl 5.7.2006