1 / 27

Analysis of CC Event Tracking Efficiency and PID Method Comparison in Recent Studies

This report revisits the CC analysis utilizing the R1.9ntuples to investigate the effects of improved tracking efficiency on particle identification (PID) methods. We compare likelihood and neural network approaches for distinguishing between charged current (CC) and neutral current (NC) events, examining performance with new and historical analysis data. The efficiency improvements in tracking yield significant insights into high-y event selections and energy resolutions. The outcomes reveal a bias in CC event recognition and demonstrate the neural network's advantages in PID performance over likelihood methods.

shaman
Télécharger la présentation

Analysis of CC Event Tracking Efficiency and PID Method Comparison in Recent Studies

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CC analysis update D.A. Petyt Sep 1st 2004 • Repeat of CC analysis with R1.9 ntuples • What is the effect of improved tracking efficiency? • Alternative PID methods: likelihood vs neural net • Gallery of events passing/failing PID cuts

  2. Effect of improved tracking in R1.9 R1.9 R1.7 R1.9 tracking better; CC selection harder (more high-y events passing cuts)

  3. Effect of improved tracking - pmu R1.9 R1.7 R1.9 tracking improvements obvious in top-right plot

  4. R1.9 reco/selection effics for QEL/RES/DIS

  5. PID performance CC NC Cut at –0.4: 85% CC efficiency, 93% NC rejection

  6. Energy resolution Showers in NC events Showers in CC events Eshw=shw.ph.GeV[0]/1.23 range

  7. Visible energy distributions CC NC reco true Positive bias in CC plot

  8. Comparison of old and new 5 year plan analysis

  9. Comparison of old and new R1.7 analysis

  10. Comparison of old and new R1.9 analysis

  11. PID: comparing techniques • Looked at neural net class in ROOT (TMultiLayerPerceptron) to see how it compares with likelihood technique for separating CC and NC events • Used same variables (event length, track pulse height fraction, track ph/plane) as likelihood analysis. Only used events with evlength<50 planes (events longer than this were assumed to be CC-like) • Advantages of NN: • Correlations between variables accounted for • No binning problems • Advantages of Likelihood method: • Simplicity, transparency

  12. Comparison of PID parameters CC NC Using re-defined PID parameter: PID=p_mu/(p_mu+p_nc) Trained NN outputs a weight: ~0 for NC events, ~1 for CC

  13. Comparison of PID performance NN does better overall – thick red curve higher than black curve. Presumably this is because correlations between variables are taken into account Likelihood seems better for low E events – not entirely sure why this is at the moment… Red: NN, Black: likelihood Thick: all events, Thin: E_nu<3 GeV

  14. CC events passing cuts Cut is PID_lik>0.95

  15. m+p+p0

  16. m+n+p++p0

  17. m+p

  18. NC events passing cuts

  19. n+p++p0

  20. 2p- + n + p0

  21. Classified NC by NN 49 planes long

  22. Long NC event passing 50 plane cut

  23. 5 p in FS – leading pp~ 5 GeV

  24. CC events failing cuts

  25. m + 1 GeV p+ + p+ + p- + n

  26. m + n + p+

  27. m + n + 2p+ +p-

More Related