Using Convolutional Neural Networks to Study Neutrinos

Abstract

The study of the neutrino, a fundamental particle, is currently one of the most promising ways to search for new laws of physics. The field aims to answer open questions such as “do neutrinos and their anti-matter partner behave the same way?” and “how many types of neutrinos are there?”. Answers to these questions will impact not only our understanding of particle physics but also influence our models of how the universe evolved. To get to these answers, neutrino experiments employ detectors known as liquid argon time-projection chambers (LArTPC) which produce high-resolution images of neutrino interactions needed to look for new physics. The analysis of such data is a challenge, requiring pattern recognition algorithms which distill the images into useful physics quantities. One approach to this problem is to apply deep convolutional neural networks (CNNs), one of the many innovations from the field of deep learning. In this talk, I will discuss recent applications of CNNs to the analysis of neutrino interactions, current challenges, and future directions using CNNs and other deep learning methods.

Bio

Taritree is an assistant professor in the Physics & Astronomy department. He received his Ph.D. from Duke University and was a Pappalardo fellow at MIT. His research interests are in using the neutrino to search for new physics beyond the Standard Model. Currently, he is focused on developing new techniques in both particle detector hardware and data analysis.