GNN-based track reconstruction for MUonE experiment

Abstract

A study of a Graph Neural Network-based model for track reconstruction in the context of MUonE experiment is presented, using simulated data corresponding to the test-run MUonE detector setup. The fully three dimensional model successfully addresses both the reconstruction and particle identification challenges essential for achieving the experiment's primary physics goal. It provides significantly faster pattern recognition than classical reconstruction algorithms, while maintaining comparable efficiency and resolution.