Jan 29 – 31, 2020
University of Venda
Africa/Johannesburg timezone

Applications of Weakly-Supervised Machine Learning Techniques in the Search for New Bosons Focusing on Dilepton Final States at the ATLAS Experiment.

Jan 30, 2020, 3:20 PM
15m
University of Venda

University of Venda

Senate Chamber P/Bag X 5050 Thohoyandou, 0950

Speaker

Mr Benjamin Lieberman (University of the Witwatersrand (ZA))

Description

In the search for new physics Beyond the Standard Model, MVA techniques are used to extract specific signal from Standard Model background processes. In this study weakly-supervised machine learning techniques are developed and evaluated using the ATLAS experiment, di-lepton (e±μ∓) final state data, in the H → Sh search. These weakly-supervised techniques use labelled background data to extract an unlabelled signal. This allows the classification of signal information without restrictions based on previously defined physics. This study uses TMVA with ROOT to evaluate the effectiveness of weakly-supervised techniques when compared to fully-supervision techniques using Boosted Decision Tree (BDT), Multilayer Perceptron (MLP) and Deep Neural Network (DNN) methods.

Primary author

Mr Benjamin Lieberman (University of the Witwatersrand (ZA))

Co-author

Dr Yesenia Hernandez Jimenez (University of the Witwatersrand (ZA))

Presentation materials