Novel Applications of Image-Processing Techniques to Particle Physics
Files
Date
2015-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Perhaps the most important unknown property of the newly-discovered Higgs boson is how strongly it couples to the top quark, the heaviest known fundamental particle. This coupling is best measured by observing combined production of the Higgs with a top quark pair, a process known as ttH. This 'signal' process is predicted to be extremely rare, as it competes with other similar 'background' processes. The most prominent of these is top quark pair production in association with bottom quark pair production, commonly denoted ttbb+jets.
In late 2015, the Large Hadron Collider resumed operation at a center-of-mass-energy of 13 TeV. This high-energy, high-luminosity environment brings with it greater chances of observing rare processes such as ttH, but produces significantly more background noise. Thus, developing accurate event discrimination algorithms is paramount to the analysis of this elusive process. In recent years, advances in the field of machine learning have produced new computer learning tools, called Deep Convolutional Neural Networks (CNNs). CNNs use the raw input pixels of photographic images to determine for themselves which features best distinguish desired signal events from unwanted background events. By visualizing the data gathered from a high-energy physics detector such as the CMS detector at CERN, we can use CNNs to uncover important information about events.
The goal of this thesis is to investigate the usefulness of applying Deep Convolutional Neural Networks to the search for the ttH process. Current results show that CNNs perform as well as standard Artificial Neural Networks in this context, with the potential for future improvements.
Description
First Place, Natural and Mathematical Sciences (Denman Undergraduate Research Forum 2016)
Keywords
Deep Convolutional Neural Networks, Higgs Boson, Large Hadron Collider, CERN, ttH, Top Quark, Machine Learning