Published February 28, 2022
| Version v1
Thesis
Open
Improving the $t\bar{t}t\bar{t}$ event selection with Graph Neural Networks in multilepton final states at the ATLAS detector
Authors/Creators
Contributors
Supervisor:
Description
A study on the use of Graph Neural Networks for the selection of$t\bar{t}t\bar{t}$ events in the ATLAS detector is presented. Data used is the Monte-Carlo simulated proton-proton collision events at $\sqrt{s} = 13$ TeV. The analysis is only concerned with the same-sign multilepton channel. After optimization, GNNs achieved a performance of AUC = $0.8744 \pm 0.0017$ which is an improvement over the previous studies conducted on the same data using Boosted Decision Trees and Feedforward Neural Networks.
Files
CERN-THESIS-2021-291.pdf
Files
(29.5 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:0f53251d8f031448b7fa8e50fa661395
|
29.5 MB | Preview Download |
Additional details
Identifiers
- CDS
- 2802704
- CDS Report Number
- CERN-THESIS-2021-291
- CDS Report Number
- BONN-IB-2022-01
CERN
- Department
- EP - Experimental Physics Department
- Programme
- No program participation
- Accelerator
- CERN LHC
- Experiment
- ATLAS