General ======== Make key results available for future reference (e.g. in a very rough INT). Same for the code. For all proposed results, if time permits, repeat for PFlow and VR. Other general considerations ============================ Each plot must carry a label with the name of the jetCollection A - Tests on jet labelling definition (no extra jet cleaning) 1. Increase the list of possible labels including charm and double hadrons. 2. Extend accordingly the "correlation map" on slide 8 of [1]. 3. Extend checks of variables done e.g. on slide 10 of [1] to all off diagonal cases; keep key diagonal entry (both labelling schemes yield one single b / one single c) as a reference depending on how the correlation map is populated, skip edge cases with very low stat. 4. Always plot variables as built from cone-based track association (the same plots, for ghost-associated tracks, could be kept as a backup, if not too heavy to produce). 5. The proposed kinematic variables for 3) are OK; would keep the SV-based variables you investigated; would add, in a weak order of priority: 1. number of tracks used by IPxD, SV1, JF; 2. number of vertices found by SV1 and JF. Here you mean just the jet pt and eta (as shown in [1]) or those we listed in the mail: hadron-jet DR in a few jet pT bins hadron pt in a few jet pT bins hadron pt/jet pT =>> Both B - Tests on jet tagging efficiency (only isolated jets, DeltaR>1 w.r.t. closest) 1. In all tagging efficiency studies, always compare coherent choices, i.e.: 1. GhostTrackAssociation + GhostLabelling; 2. ConeTrackAssociation + ConeLabelling; 3. for VR, consider VR collection for cone and VRghost for ghost. 2. Plot inclusive ROCs considering all isolated jets. ==========>>>> for b-jets (with only one b), c-jets (with only one c-hadron and no b-jets) 3. Plot inclusive ROCs considering only cases in which the flavour has been coherently labeled in both options (diagonal entries in the correlation map above). ==========>>>> for b-jets (with only one b according to both schemes), c-jets (with only one c-hadron and no b-jets according to both schemes) 4. Plot jet-pT-differential efficiency for fixed-cut WPs, for: 1. three single-jet flavours, considering only cases in which the flavour has been coherently labeled in both options; 2. if not too heavy, all WPs. ====================>>>>> SEEN ?? 5. If not too heavy, repeat as a function of b-had-pT for beauty. ====================>>>>> SEEN ?? 6. If not too heavy, repeat as a function of c-had-pT for charm. ====================>>>>> SEEN ?? Fixed cut WP: would you like that we use the recommended (calibrated) values of the cuts on the final algorithm discriminating variables [or those determined in the sample under study] ? If yes, could you tell where to find them ? Working on the right samples, I would not expect a large difference; anyway lets say the officially-released WPs for DL1r. You can find a description here: https://twiki.cern.ch/twiki/bin/view/AtlasProtected/FTAGAlgorithms2019Taggers In addition, the cuts are stored in the (browsable) CDI file, see https://twiki.cern.ch/twiki/bin/view/AtlasProtected/BTagCalibrationRecommendationsRelease21#Recommendation_March_2020 Finally, more recommendations on slide 6 here: https://indico.cern.ch/event/999600/contributions/4198307/attachments/2178406/3679073/FTAG_INTRO_2021-01-26.pdf C - Tests on track association efficiency (only isolated jets, DeltaR>1 w.r.t. closest) 1. All checks to be repeated for ghost vs cone. 2. For b- and c-jets: 1. efficiency of associating tracks from b/c-hadrons to the jet vs jet pT; matched_origin_trk_jetpT_Bchild_jetpT_B ===>>> track association efficiency vs jet pT (b-jets) matched_origin_trk_jetpT_Cchild_jetpT_C ===>>> track association efficiency vs jet pT (c-jets) 2. efficiency of associating tracks from b/c-hadrons to the jet vs b-/c-hadron pT; matched_origin_trk_bHpT_Bchild_bHpT_B vs hadron pt ===>>> track association efficiency vs b-hadron pT (b-jets) matched_origin_trk_cHpT_Cchild_bHpT_C vs hadron pt===>>> track association efficiency vs c-hadron pT (c-jets) 3. efficiency of associating tracks from b/c-hadrons to the jet vs hadron-jet DR for 3 jet pt bins. matched_origin_trk_DR_bHpt_Bchild_DR_bHpt_B ===>>>> track association efficiency vs b-hadron DR and pT (b-jets) >>>> DR to be enlarged ?? matched_origin_trk_DR_bHpt_Cchild_DR_bHpt_C ===>>>> track association efficiency vs c-hadron DR and pT (c-jets) >>>> DR to be enlarged ?? matched_origin_trk_DR_jetpt_Bchild_DR_jetpt_B ===>>> track association efficiency vs b-hadron DR and jet pT (b-jets) >>>> DR to be enlarged ?? matched_origin_trk_DR_jetpt_Cchild_DR_jetpt_C ===>>> track association efficiency vs c-hadron DR and jet pT (c-jets) >>>> DR to be enlarged ?? 3. For all jets, as a function of jet pT and b-/c-hadron pT where applicable: 1. overall number of associated tracks; n_tracks_bHpt_B_assoctrks --->average total number of tracks for b-jets (which labeling ?), vs B-hadron pt n_tracks_cHpt_C_assoctrks --->average total number of tracks for c-jets (which labeling ?), vs C-hadron pt n_tracks_jetpt_B_assoctrks --->average total number of tracks for b-jets (which labeling ?), vs jet pt n_tracks_jetpt_C_assoctrks --->average total number of tracks for c-jets (which labeling ?), vs jet pt >>>> missing: n_tracks_jetpt_L_assoctrks vs jet pt >>>> missing (not first priority): n_tracks_jetpt_BB_assoctrks vs jet pt //// jet with 2 B-hadrons POSSONO ESSERE SOVRAPPOSTI Ai C2 ? 2. overall number of associated tracks from b/c decay chain; Bjet_cut_origin_truth_label_bHpT_bHpT_B_avtracks ---> average n. of tracks for all origines vs b-hadron pt (b-jets) Bjet_cut_origin_truth_label_bHpT_binN ---> fractional flavor composition for all orig. vs b-hadron pt (b-jets) Bjet_cut_origin_truth_label_pT_binN ---> fractional flavor composition for all orig. vs jet pt (b-jets) Bjet_cut_origin_truth_label_pT_pT_B_avtracks ---> average n. of tracks for all origines vs jet pt (b-jets) Cjet_cut_origin_truth_label_bHpT_bHpT_C_avtracks ---> average n. of tracks for all origines vs c-hadron pt (c-jets) Cjet_cut_origin_truth_label_bHpT_binN ---> fractional flavor composition for all orig. vs c-hadron pt (c-jets) Cjet_cut_origin_truth_label_pT_binN ---> fractional flavor composition for all orig. vs jet pt (c-jets) Cjet_cut_origin_truth_label_pT_pT_C_avtracks ---> average n. of tracks for all origines vs jet pt (c-jets) ljet_cut_origin_truth_label_pT_binN.pdf ---> fractional flavor composition for all orig. vs jet pt (l-jets) >>>>> missing jjet_cut_origin_truth_label_pT_pT_C_avtracks ---> average n. of tracks for all origines vs jet pt (l-jets) 3. fractional flavour composition of associated tracks. trk_bH_pT_origin_truth_label_IP3D_B_binN --->> fractional flavor composition of tracks in input to IP3D vs Bhadron pt (for b-jets, which labeling ??) trk_bH_pT_origin_truth_label_SV1_B_binN --->> fractional flavor composition of tracks in output from SV1 vs Bhadron pt (for b-jets, which labeling ??) >>> missing jetFitter trk_jet_pT_origin_truth_label_IP3D_B_binN --->> fractional flavor composition of tracks in input to IP3D vs jet pt (for b-jets, which labeling ??) trk_jet_pT_origin_truth_label_SV1_B_binN --->> fractional flavor composition of tracks in output from SV1 vs jet pt (for b-jets, which labeling ??) >>> missing jetFitter TO BE REPEATED FOR c-jets, for L jets only as a function of jet pt 4. Repeat, if possible, for different levels of selection (e.g. IP3D, used by SV1). OK done (to be completed) 5. In all the above studies, check the effect of track association on the definition of jet pT itself; cfr. slide 2 and 3 of [2], in particular for VR track jets. Understand whether some more fair comparison (e.g. comparing at fixed opening) could be useful in following studies. For all categories of jets, i.e. labelled as b-jet, c-jet and light ? Yes, sorry for the confusion. That simply meant that here you should include light jets. Other tasks not necessary for signing off 1. Studies focused on non-isolated jets; repeat modified versions of B and C on dense jet environments, e.g. g->bb or boosted top hadronic decays. 2. The proposed current performance with 21.1 tracking vs ultimate tracking performance is an interesting result, but beyond the scope of signing off the QT; suggest to keep it in the to-do list for more OTP, in collaboration with other students working on track selection optimization. 3. Looking at more b-tagging-sensitive track quantities (e.g. d0 significance ecc.) can be another follow-up project. [1] https://indico.cern.ch/event/1001305/contributions/4209631/attachments/2179994/3682158/MartinoCentonze_update_28012021.pdf [2] https://indico.cern.ch/event/983708/contributions/4147498/attachments/2160408/3644980/10122020-Martino_update.pdf