Reliability of hierarchical error estimators for obstacle problems

Staff - Faculty of Informatics

Start date: 26 October 2009

End date: 27 October 2009

The Faculty of Informatics is pleased to announce a seminar given by Andreas Veeser

DATE: Monday, October 26th, 2009
PLACE: USI Università della Svizzera italiana, room SI-013, Informatics building (Via G. Buffi 13)
TIME: 14.30

Adaptivity is an important technique to efficiently exploit the available computational resources.  A key tool for the adaptive solution of boundary value problems are so-called a posteriori error estimators.  These are quantities that are computable and provide information on error sources.  An estimator is reliable if it bounds the error from above and so cannot overlook error sources.
Hierarchical error estimators have been introduced in 1983 and have been used successfully for linear, nonlinear, and in particular non-smooth problems.  While their reliability is well-understood for linear problems, their theory is its infancy for non-smooth problems like obstacle problems.

This talk will report on recent progress on the reliability of hierarchical error a posteriori estimators for obstacle problems.  To this end, we review the special properties of the obstacle problems and the linear theory of hierarchical error estimators.  Then we outline a reliability proof, which was accomplished in collaboration with Carsten Graeser, Ralf Kornhuber and Qingsong Zou.

Dr. Andreas Veeser is Associate Professor for Numerical Analysis at the Technical Institute of Milan (Politecnico di Milano). He is known to the scientific community for his publications in the field of error estimates and convergence of finite elements methods.

HOST: Prof. Rolf Krause