Shu's weekly group discussion

Wednesday 17 Dec 2025, 10:00 → 13:00 Asia/Shanghai

Tsung-Dao Lee Institute/N6F-N631 - Smart Classroom (Tsung-Dao Lee Institute)

点击链接入会，或添加至会议列表：

https://cern.zoom.us/j/4042031778?pwd=N0RlZ0l3dGtEYk9zT25OWmlwVklSZz09

10:00 → 10:10 Status

Speaker: Kang Yu (TDLI, SJTU)

1. Review the perturbation theory
2. Review the Lie algebra and its representation, especially for SO(3)
3. Get a glimpse of using structure function to calculate the cross section of deep inelastic scattering for hadrons

10:10 → 10:20 Status

Speaker: Linjing Duan (TDLI, SJTU)

CEPC Zvvy nTGC

• finished plotting the 1D, 2D signal and background histograms of two key variables
• The operator Gplus shown:  with basic selection applied:
• Photon PT in 35-110 GeV, missing E in 40-140 GeV
•  
• 
• plan to perform the 1D fitting this week

10:20 → 10:30 Status

Speaker: Jun Wei Lok (TDLI, SJTU)

• Conducted a series of measurements to study the SiPM performance under the change of temperature and bias voltage.
• Data analysis is currently ongoing and additional measurements may be required

10:30 → 10:40 Status

Speaker: Zelin Yan (TDLI, SJTU)

10:40 → 10:50 Status

Speaker: Bai-Hong Zhou (TDLI, SJTU)

• Finished:
  • ssWWjj:
    • Produced all the MC23 polarised samples;
    • Began to train the ParT for polarisation discrimination;
  • Xbb Calibration:
    • Produced all the systematics samples (except Zqq process);
• Plan:
  • DarkSHINE:
    • Work with Xuliang for the ECAL clustering and trigger task;
  • ssWWjj:
    • Produce the MC20+MC23 samples (may wait for the WZCR development);
    • Derive the final significance for WW-Frame;
  • Xbb Calibration:
    • Make plots for all systematics (for signal);
    • Calculate the yields for each systematic;
    • Start the systematic fitting chunk by chunk (and share the ws for cross-checks);

10:50 → 11:00 Status

Speaker: Xuliang Zhu (TDLI, SJTU)

ML@FPGA:

Optimization of Conv2D: parallelization of convolution pipelining

unroll width loop by a factor of kernel size (3) can avoid memory r/w conflict

latency: 10.56us (95kHz) -> 1.58us (0.6MHz)

Ideal: ~250ns (4MHz) (unroll height loop) 

Timing Violation:

• Estimate clock period exceed target(5ns)
• optimize data type / bind storage / operations

11:00 → 11:10 Status

Speaker: Zhiyu Zhao (TDLI, SJTU)

DarkSHINE:

• Examined definitions and correlations of BDT variables. Tracker variables originate from digitization, while ECAL and HCAL variables are based on truth information. Several variable pairs exhibit very high correlation (> 0.99).
• Train using the digitized ECAL and HCAL variables and compare the performance with truth information. Try to remove variables with very high correlations.

11:10 → 11:20 Status

Speaker: Haidar Mas'ud Alfanda (TDLI, SJTU)

11:20 → 11:30 Status

Speaker: Matthias Tartarin (L2IT, CNRS/IN2P3 & TDLI, SJTU)

11:30 → 12:25 REU: Undergrad Research Discussions

11:30 Status

Speaker: Peixu Teng (SPA, SJTU)

• Have been done
  • Learn some concepts of renormalization and path integrals.
  • Check the detials of MadGraph5. Especially some of the parameters settings.
• Plan to do 
  • Find time to discuss with Xuliang and Yifan about the detials of my work.
  • QFT is not yet fully learned; comrades still need to work hard.

11:35 Status

Speaker: zhijing zhang (上海交通大学李政道研究所)

11:40 Status

Speakers: Jiacheng Shi (SPA, SJTU), Zhengyang Qian (SPA, SJTU)