Speaker
Description
Event reconstruction and particle analyses in high-energy physics (HEP) rely on effective jet-flavour tagging. This presentation introduces a novel b-jet classification method using Graph Neural Networks (GNNs), which are adept at capturing complex relationships in graph-structured data. This is the first application of a GNN b-jet tagger at LHCb, aimed to enhance jet flavour-tagging through deep learning. The GNN leverages the particle identification (PID) capabilities of the LHCb detector to improve performance of the classifier. Fully-connected graphs are constructed using daughter particle information as nodes, with jet kinematics at the global level. This GNN framework is intended for further expansion with different jet architectures, allowing for more diverse applications of jet flavour-tagging.
