Dortmund 2021 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 38: Data analysis, information technology II
T 38.6: Vortrag
Dienstag, 16. März 2021, 17:15–17:30, Tm
graFEI: Full Event Interpretation using Graph Neural Networks at Belle II — •Lea Reuter1, James Kahn2, Ilias Tsaklidis3, and Pablo Goldenzweig1 — 1Institut für Experimentelle Teilchenphysik (ETP), Karlsruher Institut für Technologie (KIT) — 2Steinbuch Centre for Computing (SCC), Karlsruher Institut für Technologie (KIT) — 3Physikalisches Institut, Universität Bonn, Germany
At the Belle II experiment, flavor physics and charge parity violation are investigated at the Υ(4S) resonance. By colliding electrons and positrons, the Belle II experiment ensures a clean collision environment, where the initial state is fully defined. Neutrinos or other missing particles that cannot be directly detected, can therefore be identified using conservation laws. To analyse such processes, it is necessary to reconstruct the full Υ(4S) decay process.
The currently used Full Event Interpretation algorithm utilises Boosted Decision Trees to reconstruct the decay processes step-wise, and is therefore heavily dependent on the previous steps. The decays must be explicitly defined, which restricts the branching fraction coverage of the algorithm.
Recent works have explored Graph Neural Network approaches, since a natural representation of a decay process is a tree graph. Given their success, this work explores their application to Belle II and integrating them into the analysis software framework. Ultimately, the aim is to apply the Graph Neural Network approach to the measurement of missing energy decays.