Enhancements to the IceCube Extremely High Energy Neutrino Selection using Graph & Transformer Based Neural Networks

TL;DR

This paper introduces novel Graph Neural Network and Transformer-based neural networks to improve background rejection and directional reconstruction in IceCube's high-energy neutrino detection, enhancing sensitivity above 10 PeV.

Contribution

It presents the first application of GNN and Transformer models for neutrino classification and reconstruction in IceCube, surpassing traditional likelihood-based methods.

Findings

Enhanced atmospheric muon rejection efficiency.

Improved directional reconstruction accuracy.

Potential for better high-energy neutrino detection sensitivity.

Abstract

KM3NeT has recently reported the detection of a very high-energy neutrino event, while IceCube has previously set upper limits on the differential neutrino flux above 100 PeV but has yet to observe a neutrino event with an energy comparable to that of the KM3NeT detection. To improve diffuse measurements above 10 PeV, we apply machine learning techniques to enhance atmospheric muon background rejection and directional reconstruction. We utilize a Graph Neural Network (GNN) to perform a classification task that distinguishes neutrinos from high-energy atmospheric muons. The method allows for the rejection of early hits from laterally spread, lower-energy muons in cosmic ray showers without relying on directional reconstruction as a prior. Additionally, a Transformer-based Neural Network is implemented for directional reconstruction. Unlike previous likelihood-based rapid reconstruction algorithms that assume a single muon track, this method makes no prior assumptions about event topology of the particle inside the detector. We demonstrate improved background rejection and reconstruction performance using machine learning techniques. Applications to the development of future Extremely High Energy (EHE) selections are also discussed.