Karlsruhe 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 65: Trigger+DAQ 2

T 65.6: Vortrag

Mittwoch, 6. März 2024, 17:15–17:30, Geb. 30.23: 3/1

ML Pile-up Rejection at the ATLAS High Level Trigger (HLT) based on Hit-level Information — •Mathias Backes — Kirchhoff-Institut für Physik Heidelberg

The rejection of pile-up in proton collisions at the LHC is crucial for analyses like the Di-Higgs. One possibility to reject pile-up is by reconstructing the primary vertex position of several jets in the same event, which is implemented offline as well as in the trigger. In the case of the ATLAS experiment, this is implemented in the High Level Trigger (HLT). The main bottleneck of the HLT is the CPU usage, which is significantly increased by performing a track reconstruction.

Here we study a possibility to save computing resources by extracting the primary vertex position directly from the position of the hits in the inner detector, without performing a track reconstruction algorithm. Using ideas that have already been applied in trackless flavour tagging algorithms, we aim to construct a DeepSet neural network which is trained to predict the primary vertex z position as well as its uncertainty on a jet-by-jet basis. Events can consequently be rejected if not all required jets originate in the same vertex.

Keywords: Pile-up rejection; High Level Trigger; ATLAS; Machine Learning