Skip to content

ProbNN weights in PbPb

Carolina Aratarequested to merge
carata-PbPbProbNN_weights into master

ProbNN weights for different species in PbPb.

MC samples from sim10f for signal and background used for ProbNN training :

  • Muons: inclusive Jpsi, mm (24142001) and inclusive Upsilon, mm (18112001)
  • Pions: minimum bias (30000000), taken from Ks->pipi ; Lc and D* to enlarge the sample
  • Kaons: D*->D0pi, Kpi (27163003)
  • Protons: Lc->pKpi (25103000)
  • Ghosts are identified as MC_TRUE_ID == 0

For example, signal for muons training -> muons from J/Psi and Upsilon; background for muons training -> protons, pions and kaons from the other event types listed above. The same procedure is used for the other particles training.

This MR is linked to MR Moore!5295, which modifies the protoparticles.py code to include PbPb ProbNN weights.

Training Tuple algorithm code: hlt2_globalpid_train_data_muon.py hlt2_globalpid_train_data_kaon.py hlt2_globalpid_train_data_pion.py hlt2_globalpid_train_data_proton.py hlt2_globalpid_train_data_ghost.py

Training code: hlt2_globalpid_train_muon.py hlt2_globalpid_train_pion.py hlt2_globalpid_train_proton.py hlt2_globalpid_train_kaon.py hlt2_globalpid_train_ghost.py

pT distributions of signal and background MC_TRUE_PT_distributionMuon_Sig_Bkg.pdf MC_TRUE_PT_distributionKaon_Sig_Bkg.pdf MC_TRUE_PT_distributionPion_Sig_Bkg.pdf MC_TRUE_PT_distributionProton_Sig_Bkg.pdf

eCalTot distributions of signal and background eCalTot_distributionMuon_Sig_Bkg.pdf eCalTot_distributionPion_Sig_Bkg.pdf eCalTot_distributionKaon_Sig_Bkg.pdf eCalTot_distributionProton_Sig_Bkg.pdf eCalTot_distributionGhost_Sig_Bkg.pdf

ROC curves as a function of eCalTot and pT NNperf_ecal_Muon_PbPb.pdf NNperf_pT_Muon_PbPb.pdf NNperf_ecal_Kaon_PbPb.pdf NNperf_pT_Kaon_PbPb.pdf NNperf_ecal_Pion_PbPb.pdf NNperf_pT_Pion_PbPb.pdf NNperf_ecal_Proton_PbPb.pdf NNperf_pT_Proton_PbPb.pdf NNperf_ecal_Ghost_PbPb.pdf

Edited by Carolina Arata

Merge request reports