Measurement of Long Baseline Neutrino Oscillations and Improvements from Deep Learning

Abstract

Especially, I want to acknowledge Evan Niner for his support, his respect and also his patience, and to Michael Baird for his wisdom and example. Whatever stage of our careers Evan and Michael will continue to be my older brother examples to follow.Naturally, a special thanks should go to my closest CVN companions, Dominick Rocco who academia misses terribly, Alex Radovic my best arguing buddy, and Adam Aurisano who is just super nice.More recently, to Micah Groh, who has taken on so much in such a short time, thank you for allowing me to get you started and for all the fun that followed. And to Diana P. Mendez, Teresa Lackey, Ryan Murphy and Shiqi Yu for taking the many batons I had been running with on NOvA. Science is a passion from which half the fun is talking about. Thank you for listening.I was lucky to learn important lessons from many of my peers, for those lessons I thank Erika Catano-Mur, an unexpected companion at crunch time and an accomplice to always sneak in "just one more improvement", as well as my friends Kuldeep Kaur Maan, Biswaranjan Behera, and Liudmila Kolux paeva who are my examples of hard work and commitment and who make me always want to do better.I would also like to acknowledge people outside of NOvA who gave me advice, friendship, and opportunities for growth, such as Joseph Zennamo with whom I am delighted to talk about any physics, any day. To Sergei Gleyzer for introducing me to new perspectives to problems and the whole field, and especially my good friend Francesco Capozzi with whom I endured the longest, best hikes ever, and who in turn endured with me the proof-reading of my theory chapter.Lastly, a special acknowledgement to the people who have marked my PhD and my last few years, and potentially the few who have continued to read this far:To my friend, colleague and co-watchdog leader Louise Suter, whom I love not only for her friendship but for her example of brilliance and integrity.To Gavin Davies for being for me, like for NOvA, the string tying all loose ends together. Sometimes things seem to stand tall on their own, but there is always a support that keeps them from falling.To Justin Vasel, who is my family away from family, my brother and partner in adventure. Thank you for all the proof-reads, all the encouragement, and for always believing in me. I would not be where I am now if I didn't have you in my life.Thank you to Fermi National Accelerator Laboratory, a home away from home where I learned the true meaning and value of great American science.

New constraints on oscillation parameters fromνeappearance andνμdisappearance in the NOvA experiment

“…The coefficients of this polynomial are fit to minimize the predicted neutrino energy residuals in selected simulated ν e CC events. Whether a prong is considered electromagnetic or not is determined by a deep-learning single particle classifier that utilizes both information from the prong itself and the full event [42]. This results in an estimator with 11% resolution for both appearance signal and beam background ν e CC events in both detectors.…”

Neutrino Mass Ordering Obscured by Nonstandard Interactions

“…The dataset for the appearance channel is divided into three samples. Two of them are based on the particle identification variable (low-PID and high-PID) introduced in the context of the convolutional neural network technique for event classification [49,50]. The third one is the "peripheral" sample.…”

Context-enriched identification of particles with a convolutional network for neutrino events

“…Studies of detector response have also been implemented using this technique for the selection of photon pairs consistent with coming from neutral pion decays. The event selection using this technique achieves a purity of 92%, an improvement of 60% for the same efficiency compared to previous methods [27].…”

“…The network architecture is a simplified tower structure compared to GoogLeNet [26] and to our previous implementation [4]. The three consecutive inception modules per tower in the original architecture were reduced to one per tower [27], reducing the number of required computations with a roughly equivalent accuracy as the original classifier. The kernel size was optimized for our running time requirements, and the width of the network was reduced by a factor of two, without significant losses in performance.…”

“…This context-enriched CNN technique for particle classification has already been incorporated in NOvA's analyses, specifically in reconstructing the energy of electron neutrino interactions [27], an essential part of NOvA's primary neutrino oscillation physics measurement. The energy in electron neutrino candidate events is reconstructed from electromagnetic activity by final-state electrons and photons, and from hadronic activity by finalstate protons and pions.…”