Florian List: Machine-learning approaches to the Galactic Centre Excess

Tuesday 3 Mar 2026, 13:15 → 14:00 Europe/Stockholm

FB42 (AlbaNova Main Building)

The Galactic Centre Excess (GCE) in Fermi-LAT γ-ray data remains one of the most intriguing astrophysical anomalies, with possible origins ranging from annihilating dark matter to a population of unresolved millisecond pulsars. Traditional template-fitting techniques rely on approximate likelihoods and have been shown to produce biased results in the presence of mismodelling or in the ultra-faint point-source regime. 
In this talk, I will present deep-learning approaches designed to address these challenges. These methods bypass the need for analytic likelihoods, directly capture the statistical degeneracy between faint point sources and smooth Poisson emission, remain robust under systematics such as mild diffuse mismodelling, and jointly exploit spatial and spectral information. Applied to Fermi data, our analyses find that the GCE seems to originate from a very large population of ~10⁴–10⁵ faint sources - far beyond earlier expectations - whereas the alternative is the presence of a substantial smooth, dark-matter–like emission component.

About the speaker:  I am a Postdoctoral Fellow at the Max Planck Institute for Astrophysics, working at the intersection of cosmology, high-energy astrophysics, and machine learning. My research develops numerical methods and simulation-based inference techniques to extract fundamental physics from cosmological and gamma-ray observations. I received my PhD in Astrophysics from the University of Sydney in 2021.