KamNet: An Integrated Spatiotemporal Deep Neural Network for Rare Event Search in KamLAND-Zen

Abstract

Rare event searches allow us to search for new physics at energy scales inaccessible with other means by leveraging specialized large-mass detectors. Machine learning provides a new tool to maximize the information provided by these detectors. The information is sparse, which forces these algorithms to start from the lowest level data and exploit all symmetries in the detector to produce results. In this work we present KamNet which harnesses breakthroughs in geometric deep learning and spatiotemporal data analysis to maximize the physics reach of KamLAND-Zen, a kiloton scale spherical liquid scintillator detector searching for neutrinoless double beta decay ($0\nu \beta \beta$). Using a simplified background model for KamLAND we show that KamNet outperforms a conventional CNN on benchmarking MC simulations with an increasing level of robustness. Using simulated data, we then demonstrate KamNet's ability to increase KamLAND-Zen's sensitivity to $0\nu \beta \beta$ and $0\nu \beta \beta$ to excited states. A key component of this work is the addition of an attention mechanism to elucidate the underlying physics KamNet is using for the background rejection.

arXiv comment12 pages, dual submission with upcoming KamLAND-Zen 800 main result

References in corpus (14)

• First Results from KamLAND: Evidence for Reactor Anti-Neutrino Disappearance
• Neutrino Physics with JUNO
• Search for Majorana Neutrinos near the Inverted Mass Hierarchy Region with KamLAND-Zen
• Precision measurement of the 7Be solar neutrino interaction rate in Borexino
• New results on solar neutrino fluxes from 192 days of Borexino data
• Search for the Majorana Nature of Neutrinos in the Inverted Mass Ordering Region with KamLAND-Zen
• Experimental evidence of neutrinos produced in the CNO fusion cycle in the Sun
• The event generator DECAY4 for simulation of double beta processes and decay of radioactive nuclei
• A new measurement of antineutrino oscillation with the full detector configuration at Daya Bay
• Search for invisible modes of nucleon decay in water with the SNO+ detector
• Measurement of the ${}^8$B Solar Neutrino Flux in SNO+ with Very Low Backgrounds
• Search for double-beta decay of 136Xe to excited states of 136Ba with the KamLAND-Zen experiment
• Suppression of Cosmic Muon Spallation Backgrounds in Liquid Scintillator Detectors Using Convolutional Neural Networks
• Multi-site Event Discrimination in Large Liquid Scintillation Detectors

Cited by in corpus (5)

• Search for the Majorana Nature of Neutrinos in the Inverted Mass Ordering Region with KamLAND-Zen
• Search for Majorana Neutrinos with the Complete KamLAND-Zen Dataset
• Search for Two-neutrino Double-Beta Decay of ${}^{136}\mathrm{Xe}$ to the $0_1^{+}$ excited state of ${}^{136}\mathrm{Ba}$ with the Complete EXO-200 Dataset
• The Fast Stochastic Matching Pursuit for Neutrino and Dark Matter Experiments
• Machine learning-powered data cleaning for LEGEND: a semi-supervised approach using affinity propagation and support vector machines