arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.31391physics.ins-detcs.LGhep-ex
收藏

基于深度学习的Hyper-Kamiokande实验低能量触发算法

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

University of Warwick（沃里克大学）

Katharina Lachner, Saúl Alonso-Monsalve, Benjamin Richards, Davide Sgalaberna

备注：
18 pages, 8 figures
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文针对Hyper-Kamiokande实验的低能中微子事件（<7 MeV），提出并比较了监督式神经网络和基于异常检测（自编码器与MPDR）的触发算法，在3 MeV单电子信号上效率分别达76.7%和31.8%，远超传统命中计数触发的26.4%，且GPU推理延迟低于毫秒级，满足实时运行需求。

摘要：

现代机器学习技术因其强大的模式识别能力在粒子物理学中变得越来越重要，包括在具有严格运行时间约束的实时数据采集中。本文详细介绍了针对大型水切伦科夫探测器（如Hyper-Kamiokande）的低能中微子事件（低于7 MeV）的基于深度学习的触发算法的性能。展示了自定义神经网络监督分类器的性能，以及两种仅基于探测器噪声训练的异常检测方法：纯自编码器和基于流形投影-扩散恢复（MPDR）的能量模型。监督模型对动能为3 MeV的单电子信号识别效率为76.7%，显著超过了传统基于命中计数触发的26.4%的信号效率，MPDR方法也达到了31.8%。在GPU上的运行时间评估显示，每窗口推理延迟远低于毫秒量级，表明实时操作是可行的。

英文摘要：

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 a model based on Manifold Projection-Diffusion Recovery. 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%, while the Manifold Projection-Diffusion Recovery approach reaches 35.4% at the same operating point. Runtime evaluations on GPU yield per-window inference latencies well below the millisecond scale.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑