Daniel Herbst

I am a PhD student in ML at Technical University of Munich under the mentorship of Prof. Stefanie Jegelka. Before starting my PhD, I earned an MSc in Mathematics at TU Munich, where I worked with the DAML group on symmetries and long-range interactions in graph neural networks (GNNs) as a research assistant, and was advised by Stefanie Jegelka for my Master thesis on transferability of GNNs. During my MSc, I also completed various industry internships as well as an exchange at University of Waterloo. Prior to this, I obtained a BSc in Mathematics at Karlsruhe Institute of Technology.

My current research interests lie in the theory of machine learning, especially the interplay between weight-space geometry, parameter symmetries, training dynamics, and scaling. I am also interested in the theoretical foundations of graph learning and, more broadly, in principled approaches to learning with structure.

I am always open to collaborations and thesis/project supervisions (math or CS)—feel free to reach out!