深度学习在高能核物理中的应用  

Applications of deep learning in high energy nuclear physics

作　　者：王凌霄 庞龙刚 周凯 WANG LingXiao;PANG LongGang;ZHOU Kai(Frankfurt Institute for Advanced Studies,Frankfurt am Main 60438,Germany;Institute of Particle Physics,Key Laboratory of Quark and Lepton Physics(MOE),Central China Normal University,Wuhan 430079,China)

机构地区：[1]法兰克福高等研究院,法兰克福60438 [2]华中师范大学粒子物理研究所夸克与轻子教育部重点实验室,武汉430079

出　　处：《中国科学：物理学、力学、天文学》2022年第5期16-33,共18页Scientia Sinica Physica,Mechanica & Astronomica

基　　金：德国联邦教育与研究部ErUM-Data Project、FIAS AI Grant of SAMSON AG,Frankfurt、Xidian-FIAS联合研究中心项目;NVIDIA Corporation的GPU Grant资助项目。

摘　　要：深度学习作为一种从海量多维数据中自动提取内在关联的方法,近年来在图像处理和语义识别等领域取得了突破性的进展.这种数据驱动的计算方法,正在高能核物理的研究中逐渐崭露头角.高能核物理实验观测主要来源于小尺度高能标的相对论重离子碰撞(Heavy-Ion Collisions),同时基于第一性原理的格点量子色动力学(Lattice QCD)计算也提供了数量可观的可靠数据,而如何从中提取物理信息或优化计算成为深度学习应用的焦点.基于此,本文首先介绍如何利用神经网络在高能重离子碰撞中提取相结构和关键物理过程信息,然后再聚焦于生成算法与格点QCD计算的结合.最后介绍物理驱动的深度学习方法,并讨论这种新的研究范式可能会对本领域产生的深远影响.Deep learning methods are exquisitely tailored to uncover the structure in complex data and efficiently describe it with a finite number of parameters.This data-driven method has recently emerged in high-energy nuclear physics.Experimental observations of high-energy nuclear physics are mainly derived from relativistic heavy-ion collisions(HICs)at the small length high-energy scales,and lattice quantum chromodynamics(QCD)from the first-principle calculations,based on which the extraction of physical information or optimizing calculations has become the foundation of deep learning applications.Given this,this paper will first introduce the use of neural networks to extract the phase structure and key physical information in HICs and then focus on the combination of generative algorithms and lattice QCD calculations.Finally,a physics-driven deep learning method is presented,and the possible far-reaching impact of this new research paradigm on the field is discussed.

关 键 词：深度学习 生成模型 QCD相变 相对论重离子碰撞 格点QCD 

分 类 号：O571[理学—粒子物理与原子核物理] TP18[理学—物理]