SMuK 2023 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 62: DAQ NN/ML – GRID II

T 62.1: Vortrag

Mittwoch, 22. März 2023, 15:50–16:05, HSZ/0301

Development of machine-learning based topological algorithms for the CMS level-1 trigger — •Finn Labe, Johannes Haller, Gregor Kasieczka, Artur Lobanov, and Matthias Schröder — Institut für Experimentalphysik, Universität Hamburg

At the CMS experiment, a two-level trigger system is used to decide which collision events to store for later analysis. Due to the large fraction of non-interesting, low-energy collisions, currently used triggers often rely on momentum thresholds, only selecting events containing at least one highly-energetic object. In many cases, such as searches for di-Higgs production, this can substantially reduce the signal efficiency of the trigger selection. Targeting the upgraded CMS detector for the High Luminosity LHC, novel techniques are presented that utilize machine learning inside the first hardware layer of the trigger, which is based on FPGAs. Instead of individual objects, these triggers rely on the full event topology to select previously inaccessible events. The usage of these algorithmns in the context of the second trigger layer and offline analysis is studied.