Interpretable machine learning approach for electron antineutrino selection in a large liquid scintillator detector

TL;DR

It is demonstrated, using the JUNO detector as an example, that, despite the already high efficiency of a cut-based approach, the presented ML model can further improve the overall event selection efficiency and allows for the retention of signal events at the detector edges that would otherwise be rejected because of the overwhelming amount of background events.

Abstract

Several neutrino detectors, KamLAND, Daya Bay, Double Chooz, RENO, and the forthcoming large-scale JUNO, rely on liquid scintillator to detect reactor antineutrino interactions. In this context, inverse beta decay represents the golden channel for antineutrino detection, providing a pair of correlated events, thus a strong experimental signature to distinguish the signal from a variety of backgrounds. However, given the low cross-section of antineutrino interactions, the development of a powerful event selection algorithm becomes imperative to achieve effective discrimination between signal and backgrounds. In this study, we introduce a machine learning (ML) model to achieve this goal: a fully connected neural network as a powerful signal-background discriminator for a large liquid scintillator detector. We demonstrate, using the JUNO detector as an example, that, despite the already high efficiency of a cut-based approach, the presented ML model can further improve the overall event selection efficiency. Moreover, it allows for the retention of signal events at the detector edges that would otherwise be rejected because of the overwhelming amount of background events in that region. We also present the first interpretable analysis of the ML approach for event selection in reactor neutrino experiments. This method provides insights into the decision-making process of the model and offers valuable information for improving and updating traditional event selection approaches.

Figures (12)

• Figure 1: Reconstructed energy as a function of volume ($R^3$) for radioactivity (left) and IBD events (right). The IBD prompt energy spectrum extends up to approximately 12 MeV, while radiogenic events dominate the low energy range. The FV cut is indicated by the dashed line. The secondary axis provides the linear scale.
• Figure 2: Distributions of features for the dataset used to train and evaluate the ML model for both IBD (red) and accidental (blue) events.
• Figure 3: The schematic view of a neuron --- the basic component of a neural network.
• Figure 4: Network architecture after the optimization procedure. The 10 features introduced in \ref{['sec:problem_statement']} are used as input for a fully connected neural network with 3 layers: the input layer with 96 neurons and 2 hidden layers of 240 neurons. As an activation function for the neurons, we use ReLU functions for all the layers except for the output one where with the sigmoid function is used. Binary cross-entropy bib:logloss is used as a loss function, and Adam is used as an optimizer. The model consists of 84k trainable parameters. Being small and compact, the model can provide predictions for more than 1M events per second.
• Figure 5: Left: Medians of the metrics by solid lines and their standard deviations after the bootstrap procedure as a function of the threshold (T-value). The best threshold value is shown with the dashed line. Right: Output score provided by the FCNN model for events from the testing dataset. Most of the events are perfectly separated. The dashed line shows the best T-value. The inset plot represents a confusion matrix of the predictions.