Identification of highly boosted H → γγ decays with the ATLAS detector using deep neural networks

This thesis introduces two jet tagging algorithms to identify highly boosted H → γγ decays using the ATLAS detector at the LHC. Based on the Deep Neural Network (DNN) architecture, the first algorithm's performance is comparable to an existing algorithm designed for highly boosted Z → e+e− decays. The DNN jet tagger is also multifunctional and highly effective for identifying Z → e+e− decays. Notably, it displayed enhanced rejection rates for background (τ)τ-jets. The second algorithm leverages an Adversarial Neural Network (ANN) architecture for mass-decorrelated classification. While it exhibited a slight performance decrease compared to the DNN- based tagger, it demonstrated a 27.8% reduction in mutual information between the mass feature and scalar discriminant metric, substantiating its capability for mass-decorrelated jet identification.