Bonn 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 58: Higgs: associated production
T 58.2: Vortrag
Mittwoch, 1. April 2020, 16:45–17:00, H-1.002
Application of Deep Neural Networks to Combinatorial Assignment of Jets in a ttH(bb) Analysis in CMS — •Tobias Lösche1, Lisa Benato1, Gregor Kasieczka1, Alessandro Calandri2, Mauro Donega2, Alejandro Gomez Espinosa2, Maren Meinhard2, Christina Reissel2, Daniele Ruini2, and Rainer Wallny2 — 1Institut für Experimentalphysik, Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg — 2Insitute for Particle Physics and Astrophysics (IPA), ETH Zuerich
A precise determination of the interactions of the Higgs boson with other SM particles is a crucial part of the LHC physics program. When determining the top Yukawa coupling in ttH(bb) events, deep learning plays an integral role. In the single-lepton channel, multivariate approaches using deep neural networks (DNNs) achieve state-of-the-art performance in signal/background classification.
A particular challenge of this analysis is the discrimination of ttH(bb) events from the irreducible tt + bb background. Considering the combinatorial assignment of jets offers a possible means to deal with this problem and thus further improve performance. To achieve this, multiple DNN architectures were analyzed: An attention-based classifier, able to focus on the different combinations of objects in the event and a graph-based network, inferring relations between objects by learning a meaningful measure of distance between their respective nodes. The results of these analyses will be presented in this talk.