Study of H→WW→eμ for quantum entanglement with the ATLAS detector at CERN

This Bachelor thesis adresses the problem of reconstruction of kinematic properties of W bosons in H → WW → lνlv decay. The primary focus is to examine the capability of several machine learning algorithms to accurately reconstruct the 4-momenta of the W boson. Using simulated data with the kinematics of the final decay products, this work compares the performance of models such as Random Forest, AdaBoosted Decision Trees and Deep Neural Network using several validation metrics. The training pipeline is developed using the Python programming language to ease model training and allow performing experiments on architecture and underlying methods. The research finds that a Deep Neural Network method allows for the most precise reconstruction of the W kinematics. The contribution of this work lies in emphasizing the great potencial of using machine learning methods to reconstruct the W boson kinematics of particle systems, setting a foundation for future improvements in testing the Bell inequalities and quantum tomography.