Ponente
Descripción
The High-Luminosity upgrade of the LHC will increase the collision rate by a factor of five, resulting in dense environments with dozens of overlapping interactions. Within this context, the LHCb Upgrade II and its next-generation electromagnetic calorimeter, the PicoCal, will face major challenges in the accurate energy reconstruction of photons, electrons, and neutral pions. To address these conditions, we present a novel Graph Neural Network (GNN) approach in which clusters of calorimeter cells are represented as graphs.The model learns to mitigate the pile-up contribution, outperforming standard reconstruction techniques in energy resolution.
A lightweight, attention-enhanced variant, known as GarNet, is also explored, achieving similar accuracy with up to eight times faster inference, opening the door to real-time applications in future LHC runs.
