MODUS Seminar Program
Weekly talks in the MODUS Seminar: Wednesday, 12:30-14:00, Room S102, FAN-B
Individual talks may be held via zoom or at different times, see the information, below.
Further talks will be announced as the semester progresses. To receive email notifications about upcoming talks, please subscribe to the MODUS elearning course.
Summer Semester 2026
27 May 2026, 12:30h
Luigi Vanfretti
Professor, Department of Electrical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA
Dynamic Modeling for Autonomous μ-grid Control and Resilient Operation using Modelica and OpenIPSL
10 June 2026, 12:30h
Jirka Vomlel
Senior Research Fellow, Department of Decision-Making Theory, Institute of Information Theory and Automation (ÚTIA), Czech Academy of Sciences, Prague, Czech Republic
Unimodal Tucker Decomposition of Probability Distributions
Abstract:
We propose a novel computational method for representing multidimensional probability distributions through a unimodal Tucker decomposition. Motivated by social science surveys that use Likert scales, we interpret vague response categories as fuzzy sets to better capture their inherent uncertainty. This approach utilizes an alternating least squares algorithm combined with unimodal projections to ensure that the resulting factor matrices are both interpretable and mathematically compact. By restricting these matrices to unimodal shapes, the model aligns with natural human response patterns while significantly reducing the number of parameters required for data storage. Experimental results using Czech survey data demonstrate that this constrained method provides a high-quality approximation of original data tables. Ultimately, the work offers a more efficient and readable framework for analyzing complex probabilistic models in the social sciences.
17 June 2026, 12:30h
Jonas Püschel
Wissenschaftlicher Mitarbeiter, Lehrstuhl für Numerische Mathematik, Universität Augsburg
Energy-adaptive Riemannian optimization for Kohn-Sham type problems: an overview
Abstract:
Kohn–Sham type problems form the computational core of density functional theory and give rise to large-scale, nonlinear eigenvalue problems whose accurate and efficient solution remains a central challenge in electronic structure calculations. In this talk, we provide an overview of energy-adaptive Riemannian optimization strategies for the numerical solution of Kohn–Sham type equations. This includes theoretical foundations like construction of the energy-adaptive metric for the different Riemannian manifolds, practical aspects like the gradient calculation and general choice of parameters, and challenges and limitation of energy-adaptive approaches. The talk concludes with an illustration of practical performance in numerical experiments.
24 June 2026, 12:30h
Shivam Sundriyal
Wissenschaftlicher Mitarbeiter, Lehrstuhl für Wissenschaftliches Rechnen, Universität Bayreuth
Not all bits are created equal: Mixed precision for discontinuous Galerkin methods
Abstract:
Modern hardware tells a curious story: the gap between peak floating-point throughput and sustainable memory bandwidth keeps widening, while the AI revolution has pushed GPU vendors toward ever lower precision, FP16, FP8, and now FP4. For scientific simulations, this raises a natural question: can we borrow the same tricks without giving up the stability and convergence we cannot negotiate away?
This talk explores structure-aware low precision for discontinuous Galerkin methods. In DG discretizations, the solution is represented locally by polynomial expansions. For smooth solutions, the coefficients of these expansions often decay with polynomial degree: some coefficients carry large-scale features of the solution, while others encode finer details. Standard floating-point formats ignore this hierarchy, assigning the same number of bits to every coefficient. Not all bits are created equal, and standard formats spend them as if they were.
This motivates Adaptive Spectral Block Floating Point, a degree-aware number format that allocates precision according to modal importance. The format generalizes to arbitrary polynomial degree and bit budgets, and extends to tensor-product bases in two and three dimensions. Numerical experiments show that ASBFP can preserve expected DG convergence rates while substantially reducing memory footprint, with configurations ranging from aggressive compression to FP64-like accuracy.
The main message is that mixed precision for scientific computing need not be a blunt FP64-to-FP32 replacement. When the discretization has mathematical structure, precision can be allocated where it matters most.
1 July 2026, 12:30h
Matthias Gerdts
Professor for Engineering Mathematics, Universität der Bundeswehr München
Realtime trajectory optimization and model-predictive control for mobile robotics applications
Abstract:
Automation and autonomy becomes more and more important in many applications with mobile robots.
These mobile robots often do not just follow a precomputed reference path, but they need to be able to
update their trajectories in order to react on changing environments. This typically requires a feedback
control strategy, which takes into account the current state of the robot and the environment. To this
end we employ a model-predictive control (MPC) strategy, which requires to solve (discretized) optimal
control problems repeatedly. Achieving realtime capability is mandatory and often a challenge for
the deployment on real systems. The talk provides an overview on techniques which can be used to reduce
computational times, in particular imitation learning and problem-informed machine learning.
The talk will also focus on methods for the coordination of interacting systems, which are not
necessarily cooperative. We investigate suitable solution concepts and embed them into the MPC framework.
The first approach uses generalized Nash equilibrium problems, which allow to model the
coordination of automated agents without using pre-defined priorities. The second approach couples
scheduling tasks with optimal control and leads to a bi-level optimization formulation.
Numerical experiments and case studies will be presented to illustrate the methods.
15 July 2026, 12:30h
Jakob Grießhammer
Wissenschaftlicher Mitarbeiter, Lehrstuhl für Wirtschaftsmathematik, Universität Bayreuth
Improved Branch & Cut & Price for Location Type Problems