Domain Generalization with Machine Learning in the NOvA Experiment

1. Auflage 2023
ISBN: 978-3-031-43582-9
Verlag: Springer Nature Switzerland

Buch, Englisch, 170 Seiten, Format (B × H): 160 mm x 241 mm, Gewicht: 448 g

Reihe: Springer Theses

This thesis presents significant advances in the use of neural networks to study the properties of neutrinos. Machine learning tools like neural networks (NN) can be used to identify the particle types or determine their energies in detectors such as those used in the NOvA neutrino experiment, which studies changes in a beam of neutrinos as it propagates approximately 800 km through the earth. NOvA relies heavily on simulations of the physics processes and the detector response; these simulations work well, but do not match the real experiment perfectly. Thus, neural networks trained on simulated datasets must include systematic uncertainties that account for possible imperfections in the simulation. This thesis presents the first application in HEP of adversarial domain generalization to a regression neural network. Applying domain generalization to problems with large systematic variations will reduce the impact of uncertainties while avoiding the risk of falsely constraining the phase space. Reducing the impact of systematic uncertainties makes NOvA analysis more robust, and improves the significance of experimental results.

Zielgruppe

Research

Autoren/Hrsg.

Sutton, Andrew T. C.

Fachgebiete

• Technische Wissenschaften Technik Allgemein Mess- und Automatisierungstechnik
• Naturwissenschaften Physik Physik Allgemein Theoretische Physik, Mathematische Physik, Computerphysik
• Naturwissenschaften Physik Quantenphysik Hochenergiephysik
• Naturwissenschaften Physik Quantenphysik Kernphysik
• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz Maschinelles Lernen

Inhaltsverzeichnis

Chapter 1 Neutrinos: A Desperate Remedy

Chapter 2 A Review of Neutrino Physics

Chapter 3 The NOvA Experiment

Chapter 4 Event Reconstruction

Chapter 5 The 3-Flavor Analysis

Chapter 6 A Long Short-Term Memory Neural Network

Chapter 7 Domain Generalization by Adversarial Training

Chapter 8 Conclusion

Appendix A Notation

Appendix B Electroweak Unification

Appendix C 3 Flavor Oscillations with Wave Packets

Appendix D Neutrinos From Pions

Appendix E Gradient Reversal Layer Implementation

Appendix F The Test Beam

Über Autor(innen)

I am an experimental particle physicist focusing on neutrino physics as part of the NOvA and ANNIE experiments located at the Fermi National Accelerator Laboratory (Fermilab) in Batavia Illinois, USA. After graduating cum laude from the University of North Carolina at Charlotte with scientific bachelor degrees in Mechanical Engineering and Physics, I went on to pursue my Ph.D. at the University of Virginia in Charlottesville Virginia. Under the supervision of Craig Group, I studied neutrino physics as a member of the NOvA collaboration. Putting my engineering degree to good use, I received the US Department of Energy Office of Science Graduate Student Research Award to travel to Fermilab and assist in the construction of a Test Beam experiment for NOvA. Alongside the NOvA Test Beam, I also contributed to the main 3-flavor oscillation analysis and was selected as part of the three-person writing committee to draft the paper summarizing our 2020 results (M.A Acero et al. 2022, doi: 10.1103/PhysRevD.106.032004). My graduate education culminated in the machine learning project detailed in this book, which focuses on a technique to train more robust neural networks and reduce the impact of systematic uncertainties that limit the precision of our measurements.