Model order reduction and machine learning for parametrized problems

Basic data for this talk

Information about the event

Abstract

In this talk we give an introduction to the field of model order reduction for parametrized problems. These kinds of problems play an important role in many applications where the goal is to solve high-dimensional systems for a large set of different values of the involved physical parameters. Model order reduction provides different tools in order to reduced the computational complexity by approximating the high-fidelity system in a suitable way. We focus on reduced basis methods and describe their construction as well as application in the context of parametrized PDEs. Moreover, we discuss how to achieve an additional speedup using machine learning surrogates. The interplay between the reduced basis reduced order model and the machine learning surrogate allows to construct an adaptive and certified model hierarchy, which we showcase in different multi-query scenarios.

Speakers from the University of Münster

Kleikamp, Hendrik

Professorship of Applied Mathematics, especially Numerics (Prof. Ohlberger)