Project title: Maximising LHC discovery potential with GNNs
PI: Daniela Bortoletto
The LHC's ATLAS experiment collides protons at near-light speeds, smashing them apart to produce new particles. Our goal is to discover evidence for beyond Standard Model (BSM) particles, like dark matter, to complete our theory. All BSM searches to date use only properties of each individual collision to discriminate between BSM- and SM-like collisions. The project will pioneer this approach further by upgrading to graph neural networks (GNNs): for example, by optimising the definition of collision similarity, to obtain the first sensitivity to promising but experimentally challenging dark matter models. Furthermore, GNN analysis will revolutionise `anomaly-detection’: seeking new physics without specifying one of many possible models.
