Master Thesis defense by Morten Holm

Title: From Simulated Noise to Real Neutrinos: Tracking the Elusive Cosmic Messenger

Abstract: Neutrinos are one of the great mysteries of our time, produced in a variety of interactions throughout the cosmos; they give us insight not only into the grand scale structures of our universe, such as galaxies, but also the sub-atomic scale and the interactions with our most fundamental forces.

Finding these elusive particles is difficult and require large scale structures such as the one-of-a-kind instrument known as the IceCube Neutrino Observatory which is in the shape of a cubic kilometre of ice in Antarctica. Where light from the neutrinos' interaction with matter gives us a unique opportunity to study them.

The presentation will cover a multivariate approach within the supervised learning paradigm using a Graph Convolution Neural Networks framework known as GraphNeT, to identify characteristics in a sample of a decade of real data collected by the IceCube Neutrino Observatory.

The results of the presentation will demonstrate that for low-energy events, that multi-class classification is an effective tool in the selection process of a reconstruction pipeline of real data.

The defense takes place at Auditorium M, Blegdamsvej, but can also be followed via Zoom: https://ucph-ku.zoom.us/j/8260307060