| Summary: | Investigating hypothetical particles called dark photons helps shed light on the nature of dark matter, which is one of the biggest open questions in particle physics. This thesis presents world-leading limits in searches for prompt-like and long-lived dark photons decaying into two muons, as well as other dimuon resonances, produced in proton-proton collisions and collected by the LHCb experiment at the Large Hadron Collider at CERN.
In addition, this thesis proposes various machine and deep learning techniques and their applications to particle physics: classifier bias on a continuous feature can be controlled more flexibly with a novel moment decomposition loss function than with simple decorrelation, which can enhance bump hunt sensitivity; the first high precision generative model approach to high energy physics simulation has potential to help close the gap between pledged and required resources; we developed a simple, powerful, and novel deep learning approach to vertexing, a technique to determine the location of vertices of sprays of particles, given particle tracks; the statistics chapter is concluded by a pedagogical study of using machine learning classifiers for multivariate goodness-of-fit and two-sample tests.
|
|---|