Schöfbeck, R. (2022). Recurrent neural networks for improving measurements with the top quark [Diploma Thesis, Technische Universität Wien]. reposiTUm. https://doi.org/10.34726/hss.2022.95424
LHC; CMS Experiment; Top quark; Recurrent neural networks
en
Abstract:
The Standard Model of Particle Physics (SM) is astonishingly successful in predicting fundamental particle physics. Experiments probing the Standard Model and extensions beyond the SM (BSM), are located at CERN’s Large Hadron Collider (LHC), a particle accelerator situated at the French-Swiss boarder. The Compact Muon Solenoid (CMS), one of the four big experiments at the LHC, records proton-proton collisions at a center-of-mass energy of up to 13 TeV.Data of the CMS experiment accumulated during the Run-II of the LHC, taking place from 2016 to 2018, is analysed in this thesis. This proton collision data is used to study dierent top quark processes, namely the t ̄tZ, the tWZ and the t ̄tt ̄t signal processes. The latter two have not been observed yet and are promising targets for LHC Run III, yielding possible windows to BSM phenomena. There are many challenges in finding signal process signatures in the vast amount of data the CMS experiment provides. Often other, similar signatures give rise to huge backgrounds. Within this work, recurrent neural networks were used in order to improve signal-background discrimination in the named top quark processes.Recurrent Neural Networks (RNNs), in particular Longtermshortmemory (LSTM) networks were used to separate the signal. Expected Limits were calculated and it was shown that LSTMs are ecient in slightly improving the limits of the dierent processes.