Name: 高能物理计算和软件会议
Start: 2019-05-29T08:00:00+08:00
End: 2019-06-01T18:00:00+08:00
Location: 南京大学鼓楼校区

Wednesday, 29 May
- 会议注册
Thursday, 30 May
- 会议注册
- 开幕式
  - 1
    
    领导致辞
- 物理软件(I)
  
  Convener: Dr Weidong Li (高能所)
  - 2
    
    Applications of SNiPER Software Framework
    
    This talk will cover the applications of SNiPER in different expriments of China.
    
    Speaker: 文昊黄 (山东大学)
    
    Slides
  - 3
    
    ACTS重建框架及其应用于CEPC探测器的初步进展
    
    ACTS(A Common Tracking Software)是ATLAS实验正在开发的一个系统的重建软件框架。该框架结合ATLAS实验近年来的重建软件开发经验，将几何、事例径迹模型、寻迹、径迹的外推和拟合、校准等与径迹重建相关的各个模块包装成完整的重建系统。作为一个通用的重建框架，它的具体算法与探测器的具体几何结构无关，用户提供具体的探测器几何就可以实现各个重建模块的基本方法。通过这种灵活的模式，我们可以有一套完整的体系对不同的探测器布局进行分析和优化。另外，注重算法的性能和效率，提供并行运算也是其特点之一。报告中将要对ACTS的做简要的介绍，包括它的特点、各个模块的概念和现阶段ACTS的开发进展情况，同时也会展示现阶段其在CEPC探测器上的初步应用，包括探测器几何和材料的搭建，径迹外推，卡曼滤波拟合等。
    
    Speaker: jin 张晋 (bes3 software)
    
    Slides
  - 4
    
    Vertex Reconstruction in JUNO
    
    The Jiangmen Underground Neutrino Observatory (JUNO), currently under construction in the south of China, will be the largest Liquid Scintillator (LS) detector in the world. JUNO is a multipurpose neutrino experiment designed to determine neutrino mass hierarchy, precisely measure oscillation parameters, and study solar neutrinos, supernova neutrinos, geo-neutrinos and atmosphere neutrinos. The central detector of JUNO contains 20,000 tons of LS and 18,000 20-inch as well as 25,600 3-inch Photomultiplier Tubes (PMTs). The energy resolution is expected to be 3% at 1MeV. To meet the requirements of the experiment, a vertex reconstruction algorithm which utilizes the time and charge information of PMTs with good understanding of the complicated optical processes in the LS has been developed.
    
    Speaker: Dr ZIYUAN LI (Sun-Yat-Sen University)
    
    Slides
- 10:20
  
  茶歇
- 物理软件(II)
  
  Convener: Prof. Xingtao Huang (Shandong University)
  - 5
    
    Status of CGEM offline software
    
    A Cylindrical Gas Electron Multiplier (CGEM) detector is proposed as an update of BESIII inner tracker. This talk is to describe the status of the offline software related to the CGEM detector. The content includes the simulation, clusterization, track reconstruction, calibration and alignment.
    
    Speaker: Dr Liangliang Wang (Institute of High Energy Physics (Beijing))
    
    Slides
  - 6
    
    TAG based analysis at BESIII experiment
    
    BESIII is a new detector at the upgraded BEPCII that operated in the tau-charm threshold energy region.BESIII has accumulated the world's largest J/psi, Psi(3686), Psi(3770) data samples.To reduce the CPU time of data processing procedures is very important for the experiment to get physics results efficiently with limited hardware resources. BESIII offline software system was developed based on Gaudi. All the data procesing procedures are done event by event. Since a large fraction of events are not finally used in physics analysis. TAG based analysis can enable analysis jobs do pre-selection based on event tags and allow the job only read the events which satisfy given criteria. Event data for unselected events will not be retrieved. The CPU time of analysis jobs can be reduced obviously.
    
    Speaker: Dr Ziyan Deng (IHEP)
    
    Slides
  - 7
    
    高能物理软件框架现状及讨论
    
    Speaker: Prof. Xingtao Huang (Shandong University)
- 并行计算
  
  Convener: Dr Yaodong CHENG (IHEP)
  - 8
    
    Production experience and performance study for HEP data pro-duction at HPC-TianheII
    
    The mass Monte Carlo data production is the most CPU intensive process in the data analysis of for the high energy physics. The use of large scale computational resources at HPC in China is expected to increase substantially the cost-efficiency of the processing. TianheII, the second fastest HPC in China, which used to ranks first in the TOP500. We report on the technical challenges and solutions adopted to migrate offline software to TianheII, and on the experience and measured performance for mass production of COMET and BESIII experiment.
    
    Speaker: Mr wei 郑伟 (高能所)
    
    Slides
  - 9
    
    GPU Application in JUNO
    
    The Jiangmen Underground Neutrino Observatory(JUNO) in China is a 20 kton liquid scintillator detector, designed primarily to determine the neutrino mass hierarchy, as well as to study various neutrino physics topics. Its core part consists of O(10^4) Photomultiplier Tubes (PMTs). Event reconstruction based on this large amount of PMTs will cost a lot of time, GPU parallel computing is perfectly suitable for solving this issue. It could also be utilized in Monte Carlo Simulations, Deep Learning and many other aspects of the experiment. This talk will show a few examples of GPU application in JUNO and demonstrate its huge potential for experiments with lots of PMTs.
    
    Speaker: 罗武鸣 (高能所)
    
    Slides
  - 10
    
    高能物理高性能计算讨论
    
    Speaker: Dr Jiaheng Zou (高能所)
- 15:20
  
  茶歇
- 数据与计算平台
  
  Convener: Dr Liaoyuan Dong (高能所)
  - 11
    
    中国电子离子对撞机EicC计算需求
    
    Electron Ion Collider (EIC), regarded as the ”super electron microscope”, can provide the clearest image inside of the nucleon. It is the most ideal tool to understand the internal structure of the nuclear matter, especially the quark-gluon structure of the nucleon and nuclei. Polarized EICs are the next generation ”multi-dimensional electron microscopes” that are most effective in studying the deep structure and strong interactions of particles. Based on the Heavy Ion High Intensity Accelerator Facility which is under construction since the end of 2018 in Huizhou, the IMP is proposing to build a high luminosity polarized EIC facility in China, named ”EicC”, to carry out the frontier research on nucleon structure studies. In this talk, the current status of the EicC will be presented, including the physics programs，the detector design，and the computation requirement.
    
    Speaker: Dr Yutie Liang (IMP)
    
    Slides
  - 12
    
    高能同步辐射光源科学科学数据系统设计
    
    报告围绕高能同步辐射光源需求，结合光源类设施科学数据生命周期，介绍HEPS科学数据管理与处理系统的总体设计方案。
    
    Speaker: Mr Fazhi 齐法制 (高能所)
    
    Slides
  - 13
    
    Bookkeeping & 数据集管理
    
    Bookkeeping是基于web的元数据管理系统，运用数据库技术存储和管理离线数据的元数据和作业信息，从而实现数据文件的过程数据检索和过程追踪。Bookkeeping也为用户提供了定义良好的API和Web用户界面获取数据集的集合。
    
    Speaker: Ms HongMei 张红梅 (高能所)
  - 14
    
    面向100PB级海量数据处理的存储系统：现状及未来规划
    
    高能物理计算是典型的数据密集型计算，预计在未来十年内，高能所计算中心将建成100PB级的海量存储系统，为BESIII,JUNO,LHAASO,HEPS等高能物理实验及大科学装置提供数据存储及I/O服务。报告首先将从IT系统的角度，介绍各实验数据处理和分析对存储的需求和挑战，主要包括访问性能、可扩展性、高可用和高可靠性、可管理性、分级存储及性价比、实验数据跨域访问以及数据长期保存等。然后报告将分别介绍Lustre和EOS两个主要的磁盘存储系统针对以上需求和挑战的功能设计，这些功能设计在高能所的应用现状和问题以及我们针对这些问题做的二次开发等。特别地，我们将结合实例向软件开发人员介绍在程序性能调试过程中，存储系统能够提供的性能检测工具和手段。之后，报告还将针对软件共享和长期保存、个人数据多设备统一视图等功能需求，介绍AFS,CVMFS,IHEPBOX等存储技术以及在高能所的部署应用实践。最后，我们将介绍近期对“EOS数据副本+JBOD”和“Lustre+硬件数据冗余”两种架构的性能测试结果以及对未来存储架构的设计和考虑。
    
    Speaker: Ms Lu WANG Lu (高能所)
  - 15
    
    高能物理数据管理讨论
    
    Speaker: yaodong cheng
    
    Slides
Friday, 31 May
- 机器学习(I)
  
  Convener: Ye YUAN (高能所)
  - 16
    
    Classifying the Cosmic-Ray Proton and Light Component on the LHAASO-KM2A Experiment with the Graph Neural Network
    
    Precise measurement about the cosmic-ray (CR) component knee is essential for revealing the mistery of CR's acceleration and propagation mechanism, as well as exploring the new physics. However, strictly classifying the CR species is a tough task owing to the same hadronic interaction mechanism during their impinging on the atomosphere. Recently, rapid growth has occured in the analysis of the graph-structed data with the deep learning algorithm, which is manifested the strong strength in dealing with the structured and sparse data common in the physics researching field. In this work, we leverage the Graph Neural Network (GNN) to improve the CR-Proton and light-component classification performance on the LHAASO-KM2A experiment. The LHAASO-KM2A detectors activated by the extensive air shower are constructed as the Graph, and both the spatial manifold and the lateral distribution information of the showers are utilized. We demonstrate the effectiveness of the GNN architecture, which presents its adaptive nature over the 2 orders of magnitude of the energy, and outperforms the traditional physics-based method.
    
    Speaker: JIN Chao
  - 17
    
    基于生成对抗网络的WCDA探测器模拟数据生成
    
    研究背景: 利用Geant4模拟次级粒子在LHAASO-WCDA中的响应存在的如下的困难。 1.运行时间太长次级粒子能量越高,消耗的时间越长.使用一个CPU核模拟一个能量为100TeV的次级伽马粒子需要7小时。 2.内存消耗太大高能簇射产生的海量次级粒子信息会导致内存消耗过大的问题.而且每个带电粒子在水中前进1cm都将会产生约300个切伦科夫光子.对这些海量切伦科夫光子一一进行跟踪是不现实的。现有的优化策略有薄化处理和参数化处理,缓解了以上的问题,但是会损失精度。基于深度学习的模拟方法-生成对抗网络. 作为一种新的神经网络系统,GAN(Generative Adversarial Networks),已经在图像、文本生成等领域取得了巨大的成功。其生成的数据十分逼真，连人都无法判别数据是真实的还是生成的。因此，GAN对高能物理实验数据的模拟这的复杂任务提供了一个全新的思路—不用人为地设计模拟的规则，而是让机器去学习数据的特征、寻找数据中的规律，让机器成为“专家”。这种思路避免的复杂的物理计算过程，模型训练完成后，利用神经网络的前向传播过程快速产生模拟数据，不需要额外的资源消耗。同时，高能物理海量的实验数据为应用深度学习的方法提供了帮助。方法介绍：我们训练用于实值序列生成的GAN模型进行WCDA探测器模拟。模型中使用原初粒子信息作为GAN辅助信息,包括原初粒子质量、能量、天顶角、方位角以及次级粒子信息。探测器数据变换为3通道的序列信息。每个通道的代表的物理信息为：光电倍增管id,着火时间/ns，该时间着火次数.判别器网络为双向lstm,每个时间步输入为序列的三个通道加上辅助信息。生成器为单向LSTM，每个时间步输入为随机噪声加上辅助信息。将真实序列，生成序列分别送入判别器网络，判别器网络每个时间步通过投票进行判别该时间步的输入是否为真实序列的时间步。两个网络通过不断对抗，最终达到平衡，产生最优的判别器和生成器。我们使用最终的生成器进行模拟。
    
    Speaker: Mr 韬李 (西安交通大学)
    
    Slides
  - 18
    
    Deep Learning applied to hit classification for BESIII drift chamber
    
    Drift chamber is the main tracking detector for high energy physics experiment like BESIII. Due to the high luminosity and high beam intensity, drift chamber is suffer from the background from the beam and electronics which represent a computing challenge to the reconstruction software.Deep learning developments in the last few years have shown tremendous improvements in the analysis of data especially for object classification. Here we present a first study of deep learning architectures applied to BESIII drift chamber real data to make the hit classification of the background and signal.
    
    Speaker: 沛洵龙 (高能所)
    
    Slides
  - 19
    
    基于卷积神经网络的Muon重建
    
    JUNO实验拥有丰富的物理目标，主要包括测量中微子质量等级和精确测量中微子振荡参数。中微子实验探测器装有2万吨液体闪烁体、周围排布近18000个20英寸的光电倍增管(PMT)。中微子反β衰变（IBD）事例的主要本底来源是高能宇宙μ子带来的次级散裂中子和9Li、8He等放射性同位素，IBD事例和这些本底信号模式类似，很难从物理上进行鉴别区分，但可以通过一定时间内对探测器响应进行μ子反符合来排除这些本底。而进行μ子反符合，需要精确重建μ子的径迹信息。在真实实验中，基于μ子事例的PMT位置、收集的电荷量和最快光时间的空间分布，使用深度学习的方法进行μ子径迹的重建，该方法利用穿过顶部探测器(TT)和中心探测器(CD)的μ子事例作为训练集，使用TT重建的径迹信息作为训练集的标签。但由于穿过TT和CD的事例并不能覆盖穿过CD事例的样本空间，因此提出一种旋转的方法产生覆盖全样本空间的事例作为训练集对模型进行训练。
    
    Speaker: Mr Yan Liu (Institute of high energy physical)
    
    Slides
- 10:20
  
  茶歇
- 机器学习应用及讨论
  
  Convener: Prof. Beijiang LIU (高能所)
  - 20
    
    深度学习应用介绍
    
    Speaker: Prof. Beijiang LIU (高能所)
    
    Slides
  - 21
    
    讨论
- 空间天文(I)
  
  Convener: Dr Sheng-Sen Sun (IHEP)
  - 22
    
    GECAM空间高能天文卫星的数据处理与发布
    
    GECAM（引力波暴高能电磁对应体全天监测器）是我国正在研制的空间高能天文卫星项目，由两颗相同的微小卫星组成，运行在600公里的低地球轨道，将在6 keV-5 MeV能区探测引力波高能电磁对应体、伽马射线暴、太阳耀斑和地球伽马闪等各种空间高能辐射现象。每颗GECAM卫星配备了25个伽马射线探测器（3英寸的LaBr3+SiPM方案）以及8个荷电粒子探测器（塑料闪烁体+SiPM方案）。我们将介绍GECAM以及类似的空间高能天文卫星的数据处理与发布需求，以及GECAM的数据处理和发布的初步方案，并探讨建立适用于高能天文卫星的软件框架的可能性。
    
    Speaker: Dr 少林熊 (高能所)
  - 23
    
    Introduction to GECAM GRB data analysis and quick-analysis tools
    
    In this presentation, GECAM GRB data analysis and quick-analysis tools are introduced including the motivation about these tools: analysis modules for gamma ray burst(grb),e.g. BG estimation, light curve(LC) and energy spectrum display, fit for energy spectrum, LC analysis, calculate the correlations between grb event and gravitational waves observed, and generate 2-level data production; The computational environment and installation; the software framework and progress by now.
    
    Speakers: Dr 少林熊 (高能所) , Dr 欣颖 Song (IHEP & Forschungszentrum Juelich IKP-1)
    
    Slides
  - 24
    
    LHAASO-WCDA time calibration
    
    The Large High Altitude Air Shower Observatory (LHAASO) is a multi-component experiment to study cosmic ray physics, which is under construction near Mt. Haizishan (4410 m a.s.l.) in Southwest China. The 78,000 m$^2$ Water Cherenkov Detector Array (WCDA), which is one of the main components of the LHAASO, will be able to survey the gamma-ray sky continuously in the energy range from 100 GeV to PeV. The precision of time measurement in the detector cells of WCDA determines the sensitivity to gamma ray sources, necessitating a hardware calibration system to be deployed in the WCDA. Effects such as environment temperature variations will introduce time offsets between detector cells. In order to calibrate the time offset, a technique using high statistics low energy cosmic rays has been developed. This contribution describes the technique and demonstrates that the performance meets the requirements. The high statistics of low energy cosmic rays makes it possible to calibrate the detector in real time. The performance for 1st water pool is also presented.
    
    Speaker: Dr Jinyan Liu (IHEP)
  - 25
    
    LHAASO数据处理平台
    
    Speaker: Haibo LI (高能所)
- 15:20
  
  茶歇
- 空间天文(II)
  
  Convener: Dr 少林熊 (高能所)
  - 26
    
    HXMT数据检索与发布系统
    
    我国第一个空间天文卫星硬X射线调制望远镜(Hard X-ray Modulation Telescope，简称HXMT)卫星于2018年1月正式交付使用，本报告介绍HXMT数据检索与发布系统的设计方案和技术框架。
    
    Speaker: Mr wenshuai 王文帅 (高能所)
  - 27
    
    Ali CMB Polarization Telescope Pipeline Development
    
    Different from colliders, $AliCPT$, as a ground-based Cosmic Microwave Background experiment, meets its unique challenges of data analysis, including large I/O, relative short processing procedure, large simulations, etc.
    
    Speaker: Dr Si-yu Li (IHEP)
  - 28
    
    天文卫星数据处理框架
    
    在HXMT卫星数据中心的建设中，我们开发应用了一套适用于在线数据处理系统的轻量级框架。该框架源于开源软件Py3c，这是一套基于插件方式工作的基础组件。我们在此基础上做了二次开发，加入了消息总线用于控制指令和状态反馈，且融入了天文卫星数据处理所需的工具，形成了HXMT科学数据中心的在线数据处理系统。这套框架具备灵活的可扩展性，也可以广泛适用于其他的数据处理应用。报告将介绍这套框架的系统架构、扩展方式、应用前景，并期待就其存在的问题和进一步的开发拓展的建议与同行展开交流。
    
    Speaker: Mr 建胤聂 (高能所)
  - 29
    
    空间天文数据处理讨论
- 总结与讨论
  
  Convener: Dr Weidong Li (高能所)
Saturday, 1 June
- 返程

高能物理计算和软件会议

唐仲英楼B501

南京大学鼓楼校区