Date: July 9, 2025, 16:00–17:30

Speaker: Prof. Dr. Stefan Kesselheim (Forschungszentrum Jülich)

Location:

• Zülpicher Str. 77, 50937 Cologne
• Institute of Physics II (UoC Campus Map)
• Seminar Room II

Title: Machine Learning in Science: From Supervised Learning to Manifolds and Probabilities

Abstract: Machine Learning has proven to be very successful in solving supervised tasks - telling cats from dogs or analyzing galaxy morphology based on supervised training with labeled examples. But Machine Learning can be more: For example, it can learn complex data manifolds such as the output space of (deterministic) simulations. We can train models that identify low dimensional continuous manifolds that reveal the structure of complex high-dimensional data, and build effective surrogate models - based on physical intuition and data. We can also train classes of models to represent probability distributions - either as generative models to sample high-dimensional probabilistic datasets or to model uncertainty. This talk gives a compressed overview over non-mainstream applications of Machine Learning in science, stimulating new ideas for applications.