Date of Defense
2-6-2025 2:00 PM
Location
F3, 022
Document Type
Thesis Defense
Degree Name
Master of Science in Physics
College
College of Science
Department
Physics
First Advisor
Prof. Salah Nasri
Keywords
Photon Identification, Deep Neural Networks, Machine Learning, ATLAS Detector, Large Hadron Collider (LHC)
Abstract
Photons play a crucial role in numerous analyses at the Large Hadron Collider (LHC), particularly in studies like the Higgs boson decay to two photons. Precise photon identification is essential for enhancing the sensitivity and accuracy of such measurements. This thesis focuses on the development of a machine learning (ML)-based photon identification algorithm to improve the photon identification efficiency within the ATLAS detector, using a Deep Neural Network (DNN) approach. The primary goal is to boost photon identification efficiency by using advanced neural network techniques. Traditional photon identification relies on cuts applied to shower shape variables, which can limit the effectiveness of separating prompt photons from background signals. To overcome this, a new ML-based identification algorithm using a DNN is proposed, building on previous research that demonstrates improvements in photon identification through neural networks. This work investigates the optimization of photon identification efficiency by training a DNN with shower shape variables to differentiate between prompt and background photons. The performance of the ML-based algorithm is benchmarked against the traditional cut-based approach. The data comes from Monte Carlo Simulations.
MEASUREMENT AND IMPROVEMENT OF PHOTON IDENTIFICATION EFFICIENCIES USING MACHINE LEARNING TECHNIQUES IN THE ATLAS DETECTOR AT THE LHC
F3, 022
Photons play a crucial role in numerous analyses at the Large Hadron Collider (LHC), particularly in studies like the Higgs boson decay to two photons. Precise photon identification is essential for enhancing the sensitivity and accuracy of such measurements. This thesis focuses on the development of a machine learning (ML)-based photon identification algorithm to improve the photon identification efficiency within the ATLAS detector, using a Deep Neural Network (DNN) approach. The primary goal is to boost photon identification efficiency by using advanced neural network techniques. Traditional photon identification relies on cuts applied to shower shape variables, which can limit the effectiveness of separating prompt photons from background signals. To overcome this, a new ML-based identification algorithm using a DNN is proposed, building on previous research that demonstrates improvements in photon identification through neural networks. This work investigates the optimization of photon identification efficiency by training a DNN with shower shape variables to differentiate between prompt and background photons. The performance of the ML-based algorithm is benchmarked against the traditional cut-based approach. The data comes from Monte Carlo Simulations.
