Project 1: Data-based modeling and model reduction

“Physical” modeling of real-world phenomena is performed since centuries. Although the complexity of the systems that can now be handled has increased enormously thanks to increased computing capacities, there is still a large gap between processes that can be modeled or simulated and reality - reality is often simply too complex. One solution is data-based simulation. Due to the huge amounts of data available (big data), it was initially hoped that this could completely replace physical models. However, this has not come true, among other things due to a lack of accuracy and robustness. Therefore, in recent years, first attempts to couple "physical" and "data-based" models have been done: A physical model is to be enhanced based on data using modern methods of machine learning (ML) - the model is learned with the help of the data.

In this project, the potential of such methods for real problems is to be examined and analyzed as well as implemented in an application-oriented manner.

First supervisor:

Prof. Dr. Karsten Urban, Institut für Numerische Mathematik, Universität Ulm

 

Tandem partner:

Prof. Dr. Stephan Schlüter, Technische Hochschule Ulm

 

Consulting experts:

Prof. Dr. Henning Bruhn-Fujimoto, Institut für Optimierung und OR, Universität Ulm

Prof. Dr. Hans Kestler, Institut für Medizinische Systembiologie, Universität Ulm

Prof. Dr. Kathrin Stucke-Straub, Technische Hochschule Ulm