Abstract

I present a novel approach to LHC new-physics searches, exploiting for the first time the relationships between collider events, to discriminate between SM and beyond Standard Model (BSM) physics. A graph network is built in which individual events appear as nodes, with edges between events reflecting their similarity in kinematic space. Properties of the graph nodes describe the dataset topology, providing discrimination between BSM and SM processes based more on rarity and variability in the signal than a specific model’s kinematics. Considering several definitions of similarity provides a plethora of event variables that can be used in addition to existing kinematic information to achieve good sensitivity in a prototype electroweak supersymmetry search [arXiv:1912.10625]. Possible new physics searches benefitting from this approach are discussed. In addition, the potential to attain model-independent sensitivity to new physics is explored.