Speaker
Stefano Franchellucci
(Universite de Geneve (CH))
Description
Experiments at the LHC face exceptional challenges to acquire data, with the trigger system being from the most strained ones. The ATLAS trigger employs software-based selections at a second stage, referred to as the High-Level-Trigger. Selections on jets originating from b-quarks (b-jets) figure among the most CPU intensive ones, due to the necessity of running track reconstruction algorithms. To allow low transverse energy thresholds in selections with b-jets, a special neural network was employed to provide an early filter before executing precision tracking. The network uses coarser quality tracks, and no primary vertex reconstruction. This approach leads to significantly reduced rates while maintaining high b-jet tagging efficiencies.
Author
Stefano Franchellucci
(Universite de Geneve (CH))
