Download presentation
Presentation is loading. Please wait.
Published byTomas Humphries Modified over 9 years ago
1
First look at -jet Samples using BDT e-ID Algorithm Hai-Jun Yang University of Michigan (with X. Li and B. Zhou) ATLAS egamma Meeting March 9, 2009
2
H. Yang - BDT e-ID2 Motivation To estimate the dijet and jet rejection rates using various e-ID algorithms (IsEM, Likelihood and BDT). MC samples for test include –DS108087, jet –DS105802, JF17 dijet –DS106050, Z ee signal
3
H. Yang - BDT e-ID3 MC jet samples for test samples N_electron DS108087 jet DS105802 JF17 dijets DS106050 Z ee N_events 12788723795083690 N_candidate Et>17 GeV, | |<2.5 194046896818108550 N_candidate (precuts) With EM/Track match 204412099494153 Rejection/Efficiency after precuts Rejection 9.5 Rejection 42.7 Acceptance 86.7%
4
H. Yang - BDT e-ID4 Comparison of Input Variables
5
H. Yang - BDT e-ID5 Comparison of Input Variables -jets electrons dijets
6
H. Yang - BDT e-ID6 Results of IsEM Et(jet) > 17GeV, Tight cuts - Efficiency (Z ee) = 70.9% - Rejection ( jet) = 426(±4.7%) - Rejection (jf17) = 3092(±5.9%)
7
H. Yang - BDT e-ID7 e-ID (Likelihood) log(ElectronWt/BgWt) Et(jet) > 17GeV - Efficiency = 71% - Rej( jet)=20(±1%) - Rej(jf17)=5200(±7.6%)
8
H. Yang - BDT e-ID8 e-ID (BDT_dijet) Et(jet) > 17GeV - Efficiency = 71% - Rej( jet)=591(±5.5%) - Rej(jf17)=47830(±23%) BDT trained for dijets
9
H. Yang - BDT e-ID9 BDT peaks in jet samples?
10
H. Yang - BDT e-ID10 BDT trained for dijets BDT
11
H. Yang - BDT e-ID11 BDT trained for dijets BDT
12
H. Yang - BDT e-ID12 e-ID (BDT_ jet) Et(jet) > 17GeV - Efficiency = 71% - Rej( jet)=1788(±9.6%) - Rej(jf17)=19081(±15%) BDT trained for jets
13
H. Yang - BDT e-ID13 BDT trained for jets BDT
14
H. Yang - BDT e-ID14 BDT trained for jets BDT
15
H. Yang - BDT e-ID15 BDT trained for dijets (top 10 vars) RankInput variableGini index 1Etcone20 / Et46.08% 2E2tsts1-Emins1(Emax2-Emin in LAr.1 st )8.60% 3No. of TRT hits / No. of B-layer hits6.68% 4deta1 between track and EM cluster5.21% 5Number of pixel hits4.48% 6F1(frac. of E deposited in LAr. 1 st samp)4.32% 7Ethad1/Et (E leakage in hcal. 1 st samp)3.94% 8E237 / E2773.41% 9Eta of inner track2.33% 10Number of B-layer hits2.19%
16
H. Yang - BDT e-ID16 BDT trained for jets (top 10 vars) RankInput variableGini index 1Number of B-layer hits21.49% 2 Ntrk ( R=0.3) 17.94% 3 Pt ( R=0.3) 11.41% 4Number of pixel hits11.17% 5E233 / E2775.79% 6E237 / E2774.77% 7No. of TRT hits / No. of B-layer hits4.52% 8deta1 between track and EM cluster4.49% 9Etcone20 / Et4.20% 10F1(frac. of E deposited in LAr. 1 st samp)1.97%
17
H. Yang - BDT e-ID17 Summary samples e-IDs DS108087 Jet Rejection DS105802 JF17 dijets Rejection DS106050 Z ee Acceptance IsEM (tight)426(±4.7%)3092(±5.9%)71%(±0.4%) Likelihood20(±1.0%)5200(±7.6%)71%(±0.4%) BDT (for dijets)591(±5.5%) 426(±4.7%) 47830(±23%) 27176(±17%) 71%(±0.4%) 77%(±0.3%) BDT (for jets) 1788(±9.6%)19081(±15%)71%(±0.4%)
18
H. Yang - BDT e-ID18 Backup Slides
19
H. Yang - BDT e-ID19 BDT peaks in jet samples?
20
H. Yang - BDT e-ID20 Comparison of Input Variables
21
H. Yang - BDT e-ID21 Comparison of Input Variables
22
H. Yang - BDT e-ID22 Comparison of Input Variables
23
H. Yang - BDT e-ID23 Comparison of Input Variables
24
H. Yang - BDT e-ID24 Comparison of Input Variables
25
H. Yang - BDT e-ID25 Comparison of Input Variables
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.