Dissertation/Thesis Abstract

Searches for Supersymmetry, RECAST, and Contributions to Computational High Energy Physics
by Heinrich, Lukas, Ph.D., New York University, 2019, 330; 13421570
Abstract (Summary)

The search for phenomena Beyond the Standard Model (BSM) is the primary motivation for the experiments at the Large Hadron Collider (LHC). This dissertation assesses the experimental status of supersymmetric theories based on analyses of data collected by the ATLAS experiment during the first and second run of the LHC. Both R-parity preserving theories defined within the framework of the Minimally Supersymmetric Standard Model (MSSM) as well as R-parity violating models are studied. Further, a framework for systematic reinterpretation, RECAST, is presented which enables a streamlined, community-wide, approach to the search for BSM physics through the preservation of data analyses as parametrized computational workflows. A language and execution engine for such workflows of heterogeneous workloads on distributed computing systems is presented. Additionally, a new implementation of the HistFactory class of binned likelihoods based on auto-differentiable computational graphs is developed for accelerated and distributed inference computation. Finally, to enable efficient reinterpretation, a method of estimating excursion sets of one or more resource-intensive, multivariate, black-box functions, such as p-value functions, through an information-based Bayesian Optimization procedure is introduced.

Indexing (document details)
Advisor: Cranmer, Kyle S.
Commitee: Haas, Andrew, Hogg, David, Mincer, Allen, Weiner, Neal
School: New York University
Department: Physics
School Location: United States -- New York
Source: DAI-B 80/08(E), Dissertation Abstracts International
Subjects: Particle physics
Keywords: Reinterpretation, Supersymmetry, auto-differentiable computational graphs
Publication Number: 13421570
ISBN: 978-1-392-00471-5
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy