MLS Lab · University of Stuttgart

Machine learning for science and engineering

Developing physics-aware models and neural solvers for partial differential equations, molecular systems, and simulation.

Overview

We develop new learning methods for scientific and engineering problems, including computational chemistry and fluid dynamics. We use physical and geometric structure to design model architectures, training methods, and strategies for selecting simulation data.

Conceptual illustration of blue fluid streamlines around an airfoil, with sparse sampling points.

Research topics

  • Neural solvers for partial differential equations
  • Active learning and data-efficient simulation
  • Molecular representations and interatomic potentials

For time-dependent PDEs, we design neural solvers and active learning strategies that learn from simulations and select informative training examples.

Selected publications

Information reviewed 26 September 2026 · University research topic