利用深度学习优化西藏羊八井ASγ实验数据分析

近些年，深度学习已经越来越多的应用在粒子物理和天体物理的研究中。地面宇宙线实验ICECUBE，HAWC等也在尝试在数据处理过程中逐渐引入深度学习技术。西藏羊八井ASγ实验的核心目标之一是利用高能伽马射线反推高能宇宙线的起源。实验于2019年，观测到来自北天区最强伽马射线源蟹状星云方向的最高能伽马射线辐射，能量达到450TeV；于2021年，发现首个超高能（PeV）宇宙线加速候选天体，超新星遗迹 SNR G106.3+2.7。受限于伽马射线在所有宇宙线事例中极少的占比，挑选出的类伽马射线事例中仍然存在相当比例的宇宙线核子背景。本工作意在引入深度学习技术，进一步提高伽马/强子分离程度。同时，尝试利用深度学习技术进一步优化西藏羊八井ASγ实验能量分辨及角度分辨。以上将减少后续的伽马射线物理分析的系统误差。同时有望对10至100TeV宇宙线电子进行观测。

In recent years, deep learning has been increasingly applied in the research of particle physics and astrophysics. The cosmic ray experiments are also trying to using deep learning technology in the data analysis, such as IceCube, HAWC. One of the science goals of the Tibet ASγ experiment is to use high-energy gamma rays to reverse the origin of high-energy cosmic rays. In 2019, the Tibet ASγ experiment detected the highest energy gamma-ray radiation, which came from the Crab Nebula, the strongest gamma-ray source in the northern sky, with an extremely energy of 450 TeV. In 2021, the Tibet ASγ experiment discovered the first potential very high energy (PeV) cosmic ray accelerator, the supernova remnant SNR G106.3+2.7. Limited by less proportion of gamma rays in all cosmic rays, there is still a considerable proportion of cosmic ray nuclear background in the selected gamma-ray-like events. This work will try to take the advantage of deep learning techniques, such as image recognition, to improve the gamma/hadron separation. At the same time, this work will also try to use deep learning technology to optimize the energy and angle resolution of the data of the Tibet ASγ experiment. This will minimize the systematic errors of the subsequent physics analysis. At the same time, it is expected to observe 10 to 100 TeV cosmic ray electrons.