COLLOQUIUM: Towards a Computer Vision Particle Flow by Eilam Gross (Weizmann Institute of Science)

One of the most challenging tasks in High Energy Physics is to reconstruct the particles entering the detector from the low-level detector response data. From the Deep Learning point of view, this is a set-to-set prediction task requiring multiple features and their correlations in the input data. We deploy a set-to-set neural network architecture to reconstruct particles in events containing a single jet in a fully-simulated calorimeter. The performance comparison favors a novel architecture based on learning hypergraph structure, which benefits from a physically-interpretable approach to particle reconstruction.