Thursday, September 28
Minisymposium 9: Model order reduction in PDE-constrained optimization
Time: 10:30 - 12:30
Room: L1, Building L
Organiser: Stefan Banholzer, Universität Konstanz, Dörte Jando, Universität Heidelberg, and Stefan Volkwein, Universität Konstanz
Chair: Dörte Jando, Universität Heidelberg
The optimization and control of systems governed by partial differential equations (PDEs) is important for many applications in engineering, natural and medical sciences. However, the discretization of such problems leads to very large-scale optimization problems. To accelerate the solution process different techniques of model order reduction (MOR) have been developed. In contrast to MOR for simulation the crucial point in the optimization context is that apart from the objective also its derivatives have to be approximated appropriately. This minisymposium brings together researchers to discuss new developments and applications in this field. Important topics include MOR for Hessian
approximation, the reduced basis methods for solving large-scale optimization problems arising in data assimilation, controllability preserving MOR for parametrized PDEs and adaptive trust-region proper orthogonal decomposition (TR-POD).
Speakers:
10:30 - 11:00 | Dörte Jando (Ruprecht-Karls-Universität Heidelberg) Reduced order modeling for time-dependent optimal control problems with variable initial values |
11:00 - 11:30 | Martin Grepl (RWTH Aachen) Reduced basis approximation and a posteriori error bounds for 4D-var data assimilation |
11:30 - 12:00 | Laura Iapichino (Eindhoven University of Technology) Greedy controllability of reduced-order linear dynamical systems |
12:00 - 12:30 | Nicolas Scharmacher (Universität Hamburg) Adaptive trust-region POD for optimal control of the Cahn-Hilliard equation |