Skip to content

Latest commit

 

History

History
34 lines (19 loc) · 1.15 KB

README.md

File metadata and controls

34 lines (19 loc) · 1.15 KB
Raw

Dataset

BigMuons_MuonSelection.root

After the preselection, the following cuts were applied:

  • dR LHE-Jet<0.4

    Efficiency: 0.43

  • Partons that match to the same jet => Event removed

    Efficiency (after dR cut): 0.917

  • Just the first 7 jet in pt are considered (if one parton match a discarded jet, the event is discarded)

    Efficiency (after dR+multiple mathcing cut): 0.966

JPANet

target2.png

The input features are the ones selected with the n-1,n+1 procedures (+ the WLept reco mass and btagDeepFlavCvL)

Muons $p_t,\eta,\phi$
Neutrino $p_t,\eta,\phi,m_{W_{reco}}$
Jet $p_t,\eta,\phi,m_{Top_L}$,btagDeepFlavCvL,btagDeepFlavCvB

For each event, the output is a matrix 7 x 4 (jet x classes). (The probabilities sum up to 1 along the jets)

Taking all the 7!/(7-4)! combination we take the one which has the higher product of the jet scores

Accuracy on events: 54%