深度学习在喷注淬火中的应用  

Applications of deep learning in jet quenching

作　　者：杜轶伦 PABLOS Daniel TYWONIUK Konrad DU Yi-Lun;PABLOS Daniel;TYWONIUK Konrad(Department of Physics and Technology,University of Bergen,Bergen 5020,Norway;Department of Physics,University of Oslo,Oslo 0371,Norway;Shandong Institute of Advanced Technology,Jinan 250100,China;Alstituto Nazionale Fisica Nucleare,Sezione di Torino,Torino 1-10125,Italy)

机构地区：[1]Department of Physics and Technology,University of Bergen,Bergen 5020,Norway [2]Department of Physics,University of Oslo,Oslo 0371,Norway [3]山东高等技术研究院,济南250100 [4]Istituto Nazionale Fisica Nucleare,Sezione di Torino,Torino I-10125,Italy

出　　处：《中国科学：物理学、力学、天文学》2022年第5期157-172,共16页Scientia Sinica Physica,Mechanica & Astronomica

基　　金：Trond Mohn基金(编号:BFS2018REK01);卑尔根大学(University of Bergen)基金;挪威研究委员会(Norwegian Research Council)基金(编号:255253/F50-CERN Heavy Ion Theory);Marie Sk?odowska-Curie基金(编号:754496)的欧盟远景2020研究和创新项目;挪威电子基础设施UNINETT Sigma2数据存储和高性能计算资源(编号:NS9753K,NN9753K)资助。

摘　　要：深度学习技术在喷注淬火现象的研究中有着广泛的应用.本文回顾了近年来学者们运用不同的喷注数据表示类型和神经网络结构,在重建QCD(Quantum Chromodynamics)热密介质中的喷注动量,区分真空喷注和介质喷注等方面进行的研究,并重点介绍了深度学习在预测介质喷注的能量损失,鉴别介质中的夸克喷注和胶子喷注等方面的研究.在喷注能损的预测研究中,我们展示了深度学习技术能够逐个确定高能喷注在热核物质中穿行后的能量损失程度,以及该方法推动了喷注对热核物质的层析研究.在介质中夸克和胶子喷注的分类研究中,我们发现,随着喷注能量损失的增加,分类的精度会逐渐下降.最后,展望性地讨论了夸克和胶子喷注的子结构在介质中获得的修正.Deep learning techniques have broad applications in studying jet quenching phenomena.This paper reviews the works in recent years from several scholars employing different data representations of jet samples and architectures of neural network in terms of reconstructing the momentum of jets in a hot and dense QCD medium,distinguishing the vacuum jets and medium jets,and,in particular,identifying the energy loss of jets as well as distinguishing quark-and gluon-initiated jets in a QCD medium.In the study of jet energy loss prediction,we demonstrate that deep learning techniques can identify the degree of energy loss of high-energy jets traversing hot QCD matter on a jet-by-jet basis and show that our method advances the jet tomographic study of hot QCD matter.In the study of distinguishing quark-jets and gluon-jets in the medium,we find that the classification accuracy gradually decreases as the jets lose energy.Lastly,we discuss the medium modifications of quark and gluon jet substructures in a perspective view.

关 键 词：深度学习 喷注淬火 喷注标记 喷注重建 重离子碰撞 

分 类 号：O572.243[理学—粒子物理与原子核物理] TP18[理学—物理]