Deep-learning-based low-energy trigger algorithms for the Hyper-Kamiokande experiment

2026-05-29

AI Summary

This study addresses the inefficiency of conventional trigger algorithms in large water Cherenkov detectors such as Hyper-Kamiokande for low-energy (<7 MeV) neutrino events, which hinders real-time data acquisition. To overcome this limitation, the authors propose a deep learning–based real-time triggering scheme that combines supervised classification with unsupervised anomaly detection, significantly enhancing signal capture under stringent latency constraints. Innovatively, they apply energy-based models—specifically MPDR—and autoencoders to an unsupervised triggering scenario trained exclusively on noise data, achieving sub-millisecond inference latency on GPUs. For 3 MeV single-electron events, the supervised model attains a signal efficiency of 76.7%, while MPDR achieves 31.8%, both substantially outperforming the traditional hit-counting method’s 26.4%.

Abstract

Modern machine learning techniques have become increasingly important in particle physics because of their powerful pattern-recognition capabilities, including in real-time data acquisition where stringent runtime constraints apply. This paper details the performance of deep-learning-based trigger algorithms for a large water Cherenkov detector such as Hyper-Kamiokande aimed at low-energy neutrino events (below 7 MeV). The performance of custom neural-network supervised classifiers is shown alongside two anomaly-detection approaches trained solely on detector noise: a pure autoencoder and an energy-based model based on Manifold Projection--Diffusion Recovery (MPDR). The supervised model shows signal identification efficiencies of 76.7% for single electrons of 3 MeV kinetic energy, significantly exceeding signal efficiencies obtained from a traditional hit-count-based trigger of 26.4%, as does the MPDR approach with 31.8%. Runtime evaluations on GPU yield per-window inference latencies well below the millisecond scale, indicating that real-time operation is feasible.

Problem

• low-energy neutrino
• trigger algorithm
• Hyper-Kamiokande
• real-time data acquisition
• signal efficiency

Innovation

• deep learning
• trigger algorithm
• anomaly detection
• Manifold Projection-Diffusion Recovery
• real-time inference