Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

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

T: Fachverband Teilchenphysik

T 88: Top quarks: associated production

T 88.10: Vortrag

Donnerstag, 2. April 2020, 18:45–19:00, L-4.001

Signal extraction with ANNs for tttt production in the same–sign dilepton and multilepton channels at the LHC with the ATLAS detector — Ö. Oğul Öncel, •Niklas W. Schwan, and Markus Cristinziani — Universität Bonn

Artificial Neural Networks (ANNs) have become an increasingly popular multivariate method in particle physics. They are used in a wide range of applications such as vertex reconstruction, particle identification, calorimeter energy estimation and jet tagging.
In this talk, ANNs are considered for improving the signal extraction in the tttt analysis carried out by the ATLAS collaboration in the same-sign dilepton and multilepton channels; using the data collected during 2015–2018 with 139.4fb⁻¹ integrated luminosity and at a centre-of-mass energy of 13 TeV. The tttt events are produced in proton–proton collisions at the LHC with a cross section of around 0.01pb. The dominant background processes are the irreducible contributions from ttZ and ttH production, as well as background stemming from charge misidentification and photon conversion.
The performance of the ANNs will be compared to the Boosted Decision Tree method currently being used in the analysis. Different kinds of architectures are considered such as Feedforward and Recurrent Neural Networks which can take advantage of the high jet and lepton multiplicities of the signal process. Studies on how the Neural Network distinguishes between tttt and background events will be presented.