Adeline Paiement Université de Toulon

Machine learning and deep learning methods are increasingly popular for analysing physics and astrophysics data. However, their use often faces some specific challenges, such as the low availability of annotated ground-truth data, or the interpretability of (learning) models and of their prediction results. In this talk, we will review some recent efforts in developing learning methods that address the specific challenges of (astro)physics data. These developments exploit knowledge of the physics problem and data to inform the design of learning models.