Göttingen 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 33: Data, AI, Computing, Electronics III (ML in Jet Tagging, Misc.)
T 33.3: Vortrag
Dienstag, 1. April 2025, 16:45–17:00, VG 2.101
LHCb’s neural network-based beauty trigger: Insights from Run 3 — •Nicole Schulte1, Johannes Albrecht1, Gregory Max Ciezarek2, Blaise Delaney3, and Niklas Nolte4 — 1TU Dortmund University, Dortmund, Germany — 2CERN, Geneva, Switzerland — 3Massachusetts Institute of Technology, Cambridge, USA — 4META AI (FAIR)
The quality of the LHCb beauty physics programme relies upon b-hadron selection algorithms, particularly topological b-hadron triggers. These triggers are optimized to identify b-hadron candidates by exploiting the distinctive decay topologies of b-hadrons and their characteristic kinematic properties. As the dominant contributor to the trigger selection bandwidth, topological triggers are essential for enabling a wide range of physics analyses at LHCb.
In Run 3, LHCb introduced a novel inclusive beauty trigger which incorporates Lipschitz monotonic neural networks to enhance robustness against fluctuating detector conditions and improve sensitivity to long-lived particle candidates.
This contribution presents the performance of the inclusive topological beauty trigger across diverse conditions during the 2024 data-taking period. We demonstrate the effectiveness of these topological triggers in maintaining stable performance under varying conditions and discuss the selection efficiency using well-understood decay modes. Additionally, we examine the advantages provided by the monotonicity constraints in the trigger design.
Keywords: Neural Networks; LHCb Real Data; Run 3; Trigger; Performance