Looking for dark matter in gamma rays – can machine learning help?

We live in a ‘golden’ time for studies of high-energy (HE) astrophysics as a series of satellite and ground based telescopes currently provide high-precision data, critical to answer the century old questions on the origin of cosmic rays and physics at the heart of most energetic accelerators in the Universe. This progress has also important implications for fundamental physics, providing a test on the nature of dark matter (DM) particles, one of the long-lasting problems in contemporary physics.

Analysis of these rich data sets in a comprehensive way, inducive to identification of telltale signs of new astro- and fundamental physics, however, presents a significant challenge. As one of the most promising avenues, the machine learning (ML) techniques have been developed and embraced in a number of fields that share availability and related challenges of large data sets. In HE astrophysics however, in particular in the subfield related to cosmic rays and their interactions in the Galaxy, as well as in indirect DM searches, these techniques were rarely taken advantage of, due to particular data analysis challenges (in particular, limited training samples).

In this talk I will discuss the application of ML techniques to gamma-ray Fermi-LAT data on two concrete scientific challenges: determination of the nature of the Galactic center excess and detection and classification of (faint) gamma-ray sources, both of high relevance in both, astro- and fundamental physics. Extension and challenges of such applications to next generation telescopes (e.g. the CTA) will also be outlined.