Powered by Indico v3.0.3
Minutes of Simulation and Physics Group Meeting
Time: February 27, 2017 from 16:00 to 17:30
Location: B410
Main contents:
1. Calo optimization
Hang Zhao reported their updates briefly. He said that they are working on photon cohushconnect and separation, and also preparing paper writing.
2. Tracking + vertex optimization
Fenfen gave a report about her update on dE/dx Study.
Manqi confused about the plot on slide 5, he said that the performance of resolution should became better, when the density was changed to a lower one. However, from the plot, the trend seems to be opposite.
They decided to discuss it offline.
3. Flavor tagging based on deep learning
Bingyang Zhang gave a report on deep learning. He introduced the Deep neural network (DNN) method and showed the performance of this method. Also a comparison between this method and TMVA method was shown, and the DNN method can give better results.
There were two comments about his talk:
First is about the performance of single layer (slide 12). Bingyang said that the performance of single layer is not good.
Second is about the running time of ROC and BDT, that if the DNN need much more running time than BDT. Bingyang answered that on training stage, ROC was about 5 times longer than BDT, but on usage time, the time for two methods is comparable.
4. H->mu+mu-
Zhenwei Cui gave a update on H->mu+mu-. He compared the significance, fit numbers, and signal strength of two methods (MVA and cut-count) in different channel.(slide 2)
One discussion about his talk is that in the right plot of slide 4, the background in [120,121] and [129,130] has fluctuation, which results in the bad performance in fit. Manqi suggested changing the function of background fitting to flat function and try to increase the statistic of background.
The other discussion is about the comparison with high-luminosity LHC. He said that this results can be comparable with HL-LHC, but more details are needed.
5. Conclusion
Manqi suggested that every sub-groups should have an update every time, and the contents should be summed up before the meeting.
For round table part, sub-conveners should ask to upload the reports before the meeting.