Semantic segmentation with a sparse convolutional neural network for event reconstruction in MicroBooNE
Semantic segmentation with a sparse convolutional neural network for event reconstruction in MicroBooNE
复制标题
DOI：
10.1103/physrevd.103.052012
复制
发表时间：
2021-03-26
影响因子：
5
通讯作者：
Zhang, C.
中科院分区：
文献类型：
--
作者：
Abratenko, P.;Alrashed, M.;Zhang, C.
我们介绍了一个语义分割网络SparseSSNet的性能，它提供了MicroBooNE数据的像素级分类。MicroBooNE实验使用液态Ar时间投影室来研究中微子的性质和相互作用。SparseSSNet是一种子流形稀疏卷积神经网络，它提供了用于MicroBoones nu(E)外观振荡分析的初始机器学习算法。该网络被训练成将像素分类为五个类别，这些类别被重新分类为与当前分析更相关的两个类别。SparseSSNet的输出是进一步分析步骤中的关键输入。这项技术首次用于液氮时间投影室数据，与以前使用的卷积神经网络相比，在精度和计算资源利用方面都是一种改进。在测试样本上达到的准确率为>；=99%。对于全中微子相互作用模拟，处理一幅图像的时间约为0.5秒，内存使用量为1 GB，这允许使用大多数典型的CPU工作者机器。
We present the performance of a semantic segmentation network, SparseSSNet, that provides pixel-level classification of MicroBooNE data. The MicroBooNE experiment employs a liquid argon time projection chamber for the study of neutrino properties and interactions. SparseSSNet is a submanifold sparse convolutional neural network, which provides the initial machine learning based algorithm utilized in one of MicroBooNEs nu(e)-appearance oscillation analyses. The network is trained to categorize pixels into five classes, which are reclassified into two classes more relevant to the current analysis. The output of SparseSSNet is a key input in further analysis steps. This technique, used for the first time in liquid argon time projection chambers data and is an improvement compared to a previously used convolutional neural network, both in accuracy and computing resource utilization. The accuracy achieved on the test sample is >= 99%. For full neutrino interaction simulations, the time for processing one image is approximate to 0.5 sec, the memory usage is at 1 GB level, which allows utilization of most typical CPU worker machine.