Data-driven Modeling of Complex Physical Systems

Major research interests

My research is concerned with the design, application and theoretical analysis of data-based algorithms for modeling, simulation and analysis of large-scale stochastic dynamical systems, motivated by problems in computational physics, chemistry and materials science.
The dynamics underlying complex phenomena in the above fields often share at least one of the following characteristics: (1) little or no analytical results; (2) high-dimensional state spaces; (3) complex long-time dynamics. In my research, I am addressing these issues by developing learning methods which are specifically tailored for dynamical systems with complex long term dynamics. The focus is on data-efficient model classes which are amenable to high-dimensional problems, and which lend themselves to the development of a meaningful error theory. I have demonstrated that these methods can be used to address the challenges imposed by computer simulations of molecular systems.

 

Possible research projects (not exhaustive)

• Learning hybrid models from data
  (Collaboration partners: Prof. Sager, Prof. Benner, Prof. Mostaghim, Prof. Sundmacher, hon. Prof. Stein)