Graph Variational Autoencoder for Detector Reconstruction and Fast Simulation in High-Energy Physics

Accurate and fast simulation of particle physics processes is crucial for the high-energy community. Simulating interactions with detector both time consuming computationally expensive. With its proton-proton collision energy 13 TeV, Large Hadron Collider uniquely positioned to detect measure rare phenomena that can shape our knowledge new interactions. The High-Luminosity (HLLHC) upgrade will put a significant strain on computing infrastructure budget due increased event rate levels pile-up. Simulation highenergy collisions needs be significantly faster without sacrificing accuracy. Machine learning approaches offer solutions, while maintaining high level fidelity. We introduce graph generative model provides effiective reconstruction LHC events calorimeter deposits tracks, paving way full simulation.