Fullscan ID tracking studies (qualification task)

Look Here

Tools and setup

Following packages of Athena 21.3 have been checked-out and built locally on lxplus in folder /afs/cern.ch/user/o/okovanda/public/pilsner:

Trigger/TrigAnalysis/TrigInDetAnalysis, Trigger/TrigAnalysis/TrigInDetAnalysisUser, Trigger/TrigAnalysis/TrigInDetAnalysisExample, Trigger/TrigAnalysis/TrigInDetAnalysisUtils, Trigger/TrigHypothesis/TrigBjetHypo, Trigger/TrigValidation/TrigInDetValidation

Upon login, the workspace is set up by cd into /afs/cern.ch/user/o/okovanda/public/pilsner and sourcing two scripts:

source ~/bin/release_setup.sh

source ~/bin/setup.sh 21.3,Athena,<nightly> (the nightly is printed-out by the release_setup.sh and has to be changed from time to time as nightlies become deleted)

This ensures that we're using the binaries available from the local athena build instead of the base one.

Sample

The trigger runs on data in the RDO stage of reconstruction. These were not available for our samples of interest: MC16e Bhh (run 300760), Bsmumu (run 300426) and Bd (run 300430). Therefore, their production was requested and approved on 20th April 2020 in statistics 1M events for Bs/Bd mumu and 8.5M events for Bhh.

In order to proceed in the meantime a small RDO sample of 1000 Bhh events had been reconstructed locally for testing purposes. This was done by running the reconstruction of an existing HITS file from the dataset mc16_13TeV.300426.Pythia8BEvtGen_A14_CTEQ6L1_Bs_mu3p5mu3p5.simul.HITS.e4889_e5984_s3126

The reconstruction command and tags were obtained using command: GetTfCommand.py --AMI r10201 . These tags had to be altered by changing "CONTAINER_SPLITLEVEL = \'99\'" to "CONTAINER_SPLITLEVEL = \\'99\\'" and by adding low pT and high pT minbias HITS files. For the minbias, a file from datasets

mc16_13TeV.361238.Pythia8EvtGen_A3NNPDF23LO_minbias_inelastic_low.simul.HITS.e4981_s3087_s3111 (low pT)

mc16_13TeV.361239.Pythia8EvtGen_A3NNPDF23LO_minbias_inelastic_high.simul.HITS.e4981_s3087_s3111 (high pT)

The reconstruction itself was done in Athena 21.0.53 using this script: HITS2RDO.sh and the resulting RDO file can be found in /afs/cern.ch/work/o/okovanda/public/rdoTest/outRDO.root.

This procedure was adviced (and example scripts were provided) by Kunihiro Nagano.

Running the trigger on an RDO file

Next, the tracking trigger algorithm was run on the testing RDO file. This has been achieved using a script from TrigValidation package: TrigValidation /TrigInDetValidation/test/test_trigindetvalidation_bjet_ibl_pu80.sh. This script had been modified to run on local RDO file by replacing an if- block starting on line 206 and setting the variable jobList="ARTConfig=['/afs/cern.ch/work/o/okovanda/public/rdoTest/outRDO.root']". When executing this script, flag --local was used. This, amongst other, performs the tracking and produces a track collection ntuple that can be read by TIDAreader and analyzed by TIDArdict, which will be used in the future.

Enabling fullscan tracking

In order for the tracking algorithm to scan the entire detector, a change was introduced into the process of building the SuperRoi from constituent RoIs for B-jets in the file TrigHypothesis /TrigBjetHypo/src/TrigSuperRoiBuilderAllTE.cxx. Normally, the SuperRoi is initialized and flagged as composite, i.e. containing constituents, and in a subsequent loop, RoIs of individual jets are added to the composite SuperRoi. The tracking is then performed in the SuperRoI. In order to perform a fullscan tracking, we therefore need this super RoI to span the entire ID.

At first, an attempt was carried out to force the fullscan tracking by disabling the 'composite' flag of the SuperRoI and disabling the loop that adds the individual jet RoIs to it. This resulted into RoI of zero size (and of course no tracks).

What proved as a viable option was to pass 'true' to the constructor of the SuperRoI: TrigRoiDescriptor* superRoi = new TrigRoiDescriptor ( true), since this flags the RoI as fullscan, and is false by default. In addition, the composite flag and the loop adding constituent RoIs were both disabled.

Now the tracking indeed finds tracks in the full volume of the detector in the chains:

Chain HLT_j55_boffperf_split:InDetTrigTrackingxAODCnv_BjetPrmVtx_FTF:SuperRoi

Chain HLT_j55_boffperf_split:InDetTrigTrackingxAODCnv_BjetPrmVtx_FTF:SuperRoi:xPrimVx

Chain HLT_j55_boffperf_split:InDetTrigTrackingxAODCnv_BjetPrmVtx_FTF:SuperRoi:EFHistoVtx

Analysis of FS tracking

This will be done as part of the TIDArdict program which loops over the track ntuples and events -> chains -> rois -> tracks therein.

fullScan track ntuple

*THIS IDEA WAS LATER ABANDONED, BUT THE FLAT NTUPLE IS STILL GENERATED*

For the analysis, testing and optimization purposes, it was decided to produce an ntuple containing the fullScan tracks only, and analyze it with standalone macros rather than with Athena packages. The procedure developed here can be integrated into Athena packages once completed. The first version of this "working" ntuple has been produced by running the tracking as described above, this time on an RDO produce by the MC group - file 0000004 from the mc16_13TeV.300760.Pythia8B_A14_CTEQ6L1_B_hh.recon.RDO.e7593_e5984_s3126_r10724_tid21090554_00 dataset. This file contains 2000 events out of which the hacked bjet trigger fired on 63. In total over 7000 tracks were found in the fullScan, and written into the ntuple as one track per entry. The leafs are: run number, event number, eta, phi, pT, z0, a0, chi2, dof, errors of eta, phi, pT, z0, a0. A copy of this ntuple, for the time being, resides in /afs/cern.ch/work/o/okovanda/public/qualiTask/produceNtuple.

bjet chain problems

An issue has been encountered in the bjet chian. Lowering pT cut on the pattern recognition in the triggering algorithm led, somewhat counterintuitively, to fewer rather than more tracks reconstructed by the trigger.

cutsCompare.png
All Track pT for various cuts on pattern recognition
cutsCompare-matched.png
Reference-matched Track pT for various cuts on pattern recognition
This is investigated by Mark.

We might need to use beamspot chain instead of bjet if this is not resolved.

Beamspot histat sample

Until the problems with bjet chain are resolved, we will use the beamspot chain for the fullscans. This chain does not have the internal pT cut problems. The high statistics testing sample contains 1000 events, and has been created using the Trigger/TrigValidation/TrigInDetValidation/test/test_trigindetvalidation_beamspot_ttbar_pu80_fs.sh script. The input file was the same as for the bjet chain above.

FullScanner class

A class called FullScanner was incorporated into the rmain.cxx, taking care of our stuff. It can apply cuts and combine the tracks in each event into B candidates. Invariant mass plots combining fullscans of all 1000 events are shown with visible B-peak under preliminary pT and delta Z cuts.

  RECONSTRUCTED TRACKS REFERENCE TRACKS
NO CUTS
invMass noCuts ref.png
Invariant mass spectrum without cuts, reference tracks
2 GeV track pT, 2 GeV B pT
4 GeV track pT, 4 GeV B pT
4 GeV track pT, 1 mm track delta Z, 7 GeV B pT
4 GeV track pT, 1 mm track delta Z, 7 GeV B pT, 4766 < m < 5966

More discriminating variables:

The following variables have been implemented in the FullScanner class: lead/trail track pT, track d0, B candidate charge, B candidate delta Z, B candidate pT, B candidate mass, B candidate lifetime, B candidate pointing angle (and secondary vertex position + lxy as intermediate steps)

The distributions of these in sample of trigger tracks with loose cuts pT trail 4 GeV, pT lead 6 GeV, B cand pT 2 GeV and B candidate charge = 0 are presented:

Limiting the truth tracks:

It was desirable to limit the tracks that make it to our final analysis ntuple to only those originating from a B0 decay and belonging to either a charged Pi or K (pdg id's 211, 321 respectively). This was achieved by hacking the truth selection first in AnalysisConfig _Ntuple.cxx in TrigInDetAnalysisExample which drives the ntuple building (choosing only the pi's, k's and potential photons from radiative decays, together with the parent B0). It was observed that approx. 17% of the MC events do not contain B0 at the truth level, which might be due to oscillations into B0bar and subsequent decay.

Strange shape is observed in the invariant mass distribution of the truth tracks:


BCandMass weird.png
Strange shape

UPDATE 8.9.2020: This weird shape was due to wrong masses attributed to the four-vectors of the hadrons. All of them were set to pion mass, which naturally places correctly only the B0->2Pi. This was fixed by introducing separate masses for pi and K 4-vectors.

Truth track variable distributions:

These are the above distributions resulting from truth tracks only:

The SV position distribution, that is clearly shifted to positive X, Y values, was due to the beamline correction wasn't carried out for some reason. This was later fixed by forcing the correction there, and the distribution went back to the origin:


Truth plots below this point include this correction.

Truth track variable distribution: Is our d0 sign convention opposite to what it should be?

These plots (pointing angle, lxy, decay length) would suggest that...

Same thing happens with the trigger tracks matched to these truth tracks:

The way in which the SV are obtained:

The track parameters used to parametrize the straight lines in the transverse plane (our approximation to the actual tracks) are phi and d0. The SV is obtained as an intersection of two tracks. The d0 is signed to resolve the ambiguity, as otherwise there would be two tracks parametrized by a singe (unsigned) impact parameter and phi.

The convention for signing the d0 used here is: sgn((RxpT).z), where R is the vector which we'd draw as the impact parameter, so is the pT vector, and z is the unit vector in the z direction.

The goal is to get the line parametrization based on the d0 sign, so that we can search for their intersections.


20201006 180428.jpg
Calculation

Essentially, it all boils down to this (verified on multiple d0 'positions'):

d0 > 0: line parametrization X(t) = |d0|*cos(φ - π/2) + t*cos(φ), Y(t) = |d0|*sin(φ - π/2) + t*sin(φ)

d0 < 0: line parametrization X(t) = |d0|*cos(φ + π/2) + t*cos(φ), Y(t) = |d0|*sin(φ + π/2) + t*sin(φ)

When I find the intersection of such lines, i.e. solving X1(t) = X2(s), Y1(t) = Y2(s), for, say t & plugging it into X1(t) and Y1(t), the vector pointing from the origin to the intersection is pointing mostly oppositely with respect to the (pT1 + pT2) vector.

Obtaining background sample

Decision was made to use HLT tracks obtained from running the ID trigger on jet data as a background sample to estimate the rates. This proved to be extremely difficult, if not impossible in Athena release 21.3, in which the tracks from the B->hh' MC were obtained.

After countless failed attempts with 21.3, a light at the end of the tunnel was spotted in the release 22, which had some tests set up that were supposed to produce an AOD (containing, amongst other, also the tracks). There is a job in 22 which casts the tracks from an AOD into the track ntuple. First, we had to see if the FS tracking looks the same in release 22 as it did in 21.3.

Comparing FS tracking in 21.3 and 22

The chain used in 21.3 was the beamspot chain, which is supposed to be close regarding the FS tracking features to jet chains in release 22.

The comparison between fullscans of 21.3's beamspot and 22's j45 chains is in the following presentation: FS_update-30-9-20-v2.pdf

Signal & background variable distribution

These distributions show the signal sample - trigger tracks matched to MC truth tracks originating from the B decay with appropriate masses assigned and background - in green, it's unweighted enhanced bias data (no special selection of the events) and in red the remaining trigger tracks from the B->hh' MC, that are not matched to truth. The last needs to be taken with caution, as the interactions overlaid to signal in the MC might be repeated.

There is a 4 GeV track pT cut on the trigger tracks - matched trigger tracks, prompt, and data.

   
An updated table with normalizations to all (passing and non-passing) events above the global 4GeV cut and some additional variables is here:

The numbers in the legend (not useful here yet) are the cut efficiencies - ratios of tracks passing a potential cut wrt. initial number of candidates. The 4 GeV pT cut mentioned above is implicitly assumed to be there from the very beginning, i.e. efficiency 1 is with the cut included. With track parameters, there is 2 instead of 1 without cuts, because there are two tracks per candidate. I didn't fix this, as it might even not be used eventually.

     

UPDATE: INCLUDED Bs component. That was missing, as it hasn't been extracted from the MC truth properly. Now Bs->hh' decays had been added.

   

Cut optimization & rate estimation strategy

The selection can be optimized by the following iterative procedure. Let us assume we've got a set of n discriminating variables, and a signal and background sample, corresponding to known integrated luminosities Ls and Lb. To optimize signal acceptance and background rejection, we want to pick a figure of merit that we want to maximize with our cuts. Such a figure can be the significance of signal in counting experiment, i.e. S/sqrt(S + B) where S = Npass,s/Ls and B = Npass,b/Lb, the number of events passing per unit luminosity. We can then employ the following iterative procedure:

  1. Have a set of loose cuts on each of the n discriminating variables
  2. Find best cut on each variable with cuts from previous point applied to all other variables
    1. Take first discriminating variable and find a cut that maximizes signal significance, store the value for next iteration
    2. Take second discriminating variable and find a cut that maximizes signal significance, store the value for next iteration
    3. ...
    4. Take n-th variable ...
  3. Exchange the cuts in 1st point for those found in second
  4. if ( ! converged) goto 2.
  5. Once cuts converge, the optimization is complete (here I presume the convergence is |cut_i - cut_i+1| < epsilon where cut is a vector of all the cut values, i labels iteration, epsilon is a vector of small values)

Possible implementations

Although we can now apply cuts, the current approach of extracting track parameters mainly for plotting purposes can't be utilized and further developed for the selection optimization and rate estimation mainly because this is a per-candidate approach. We will need to retain the event structure of our signal and background samples in order to be able to count the events passing our selections. The current, per-candidate workflow starting with the MC/data sample and ending with the histograms such as those above is summarized in the following diagram.


current.jpg
Current, per-candidate workflow

Since the result of the HLT tracking - the TrkNtuple contains the data structured per event, we will only need to change how those are handeled. The ntuple is processed by the TIDArdict executable, where it loops over the events, accessing the tracks they contain. The cuts can be varied through an external configuration file. We could therefore, for each cut, run the current workflow and at the output of TIDArdict collect the numbers of events passing and failing that cut. That would mean external routines modifying the configuration files and routines evaluating the output. Implementing some sort of iterative optimization of the cuts would probably be complicated.

Alternatively, we could extract the truth matching from TIDArdict and use it as part of a new class that would read in the signal and bkg files only once, and then would perform the optimization all within itself through a set of member functions, which could be a lot easier (and faster) than handling an external executable with it's own configuration file each, and reading in the files each time we change the cut. At this moment, I would prefer this alternative.

Even more alternatively, we could somehow re-introduce the by-event structure to our current by-candidate output, but that would only save the loop over TIDAevents and matching wrt. previous idea.

Sample luminosity

For this we will need to know the integrated luminosity of our signal and bkg samples, so that our passing-event counts correspond to a unit integrated luminosity.

To calculate the corresponding luminosity of a certain number of our B->hh' signal events (where only the B0 is forced to decay into hadrons), I've so far managed to come up with the following list of effects to take into account:

  • anti-b quark production cross-section
  • fd - hadronization probability of anti-b into B0
  • B0 -> hh' branching ratio
  • further, if we use only those events that contain our track pair
    • The probability of B0 to oscillate into B0bar
    • Truth-matching efficiency (for matching pairs of trigger tracks to pairs of Truth)
For the enhanced bias background, the total integrated luminosity could hopefully be known as well as the total number of events, and then for our purposes we could re-scale the total luminosity by a fraction of events we use out of the total. Moreover we will need to re-scale by the fraction of events with L1 seeds that our selection will run on.

Rate estimation

For this we need to know the total L1 rate of the union of trigger chains that our selection will run with. Assuming our rate would be largely driven by background events, and that selection will pick events uniformly in events with different L1 seed, we would only need to run it on a number of bkg events with those L1 seeds. Then, fraction of passing times the total L1 rate of these seeds would give us the rate of our selection. We will need to make sure the assumptions hold, and possibly adapt our plan accordingly.

B candidate Matching efficiency

This was tested on an MC sample of 2000 events. This sample contains 359 events that don't have B0 (it oscillated to anti B0). Therefore, at best the HLT could have seen 1641 B candidates. It is enough if only one of the tracks is not matched - then that signal event doesn't have a B candidate. The following table contains numbers of events in which both tracks belonging to the truth B0 decay were matched.

delta R cutoff Matched Not matched efficiency
0.1 1502 139 91.5 %
0.2 1580 61 96.3 %
0.3 1617 24 98.5 %
0.4 1627 14 99.1 %

<a name="Normalization"></a>Optimization sample luminosities

Signal

In the 2015/2016 B->mumu analysis, there were 2.7 +/- 1.3 B->hh' events in the top 3 BDT bins after preselection and application of muon requirements. Therefore, assuming if there were N B->hh' events in the entire dataset, we have:

N*(preselection efficiency)*(54% BDT efficiency)*(0.1 or 0.08 % fake rate for h)*(0.1 or 0.08 % fake rate for h') = 2.7 +/- 1.3

This N then corresponds to the 26.3 fb-1 of the 15/16 dataset.

There are 1497 events in the optimization sample that contain a truth matched hadron pair.

Alternatively, we'll have a look at how many events survive the analysis selection in the 300431 sample used in 15/16 analysis. From that we'll get the luminosity of the entire 300431 sample of 5M events. The resulting per-event luminosity will have to be applied to the 300760 sample used in this study. From the job options used to generate the two samples:

300431: https://gitlab.cern.ch/atlas-physics/pmg/infrastructure/mc15joboptions/blob/master/share/DSID300xxx/MC15.300431.Pythia8B_A14_CTEQ6L1_B_hh.py

300760: https://gitlab.cern.ch/atlas-physics/pmg/infrastructure/mc15joboptions/blob/master/share/DSID300xxx/MC15.300760.Pythia8B_A14_CTEQ6L1_B_hh.py

It seems there should be a different cut on the bbar quark pT - 7 GeV in 300431 vs 5 GeV in 300760. This cut is however not visible in the truth bbar distribution - seems to be 7 GeV in both samples:

These plots are comparing the pT and eta distributions of the earliest bbar leading to the B->hh' process (left), and of the bbar closest to the B->hh' process (right). In the latter, the 7 GeV cut is visible in both samples. These values are taken from the MC directly and are not boosted into a different reference frame.

histos300431.png
pT and eta distribution of bbar quark in 300431
histos300760.png
pT and eta distribution of bbar quark in 300760

<a name="lookHere"></a>It is in fact not the bbar quark that should be subjected to different cuts in the two MC samples. What they are supposed to differ in is the pTHat cut, which was not exactly straightforward to find. This is the cut on the pT of partons leaving a hard 2 -> 2 process in the CMS frame of that process. Moreover, the 4-momenta of the final state partons are corrected for mass by having the energy component set to half of the CMS energy each, and the 3-momentum components multiplied by factor E_cms/(2*|p_cms|), i.e. the ratio of the total CMS energy and it's kinetic part. This effectively makes the particles massless. I have no idea why this correction is applied, but only with this correction, the pTHat cut is sharp. The top row is without the mass correction, the bottom row does contain this correction. The left collumn shows the CMS pT wrt. the incoming particles, the right one in lab transverse direction.

histos pTHat 300431.png
pTHat distributions in 300431
histos pTHat 300760.png
pTHat distributions in 300760
The efficiency of the 7 GeV pTHat cut on the 300760 MC using the mass-corrected pTHat wrt. incoming particles is ε = 0.922 +/- 0.002 (binomial error).

Background

Bkg: full 360026 EB run in COMA: https://atlas-tagservices.cern.ch/tagservices/RunBrowser/runBrowserReport/runBrowserReport.php?runs=360026

run lumi: 505311 nb-1

run events total: 553927183

EB events total: 110755

Non-empty EB events atm used for the bkg here: 974

Therefore naively: ( (Total run lumi)/(total run events) )*(events used) = 0.889 nb-1 ?

Merit uncertainty:

What is the uncertainty on S/Sqrt(S + B)?

Optimization & cut flow

This is what can be obtained from the optimization code - an example for 3 iterations and several variables to be optimized.

cut flow: cutFlow.txt


visIter graph.png
Cut-value scans of merit function in 3 iterations

Preliminary Optimization run with not completely nonsensical normalizations

A full optimization procedure was tested with proper normalizations of signal and background. The normalizations, i.e. per-event-luminosities, are calculated here for signal: Bhh.xlsx and for background as instantaneous luminosity divided by L1 rate. The instantaneous luminosity was taken as that of the corresponding LB of data used for optimization, and L1 rate was taken as 100 kHz. Error on this term is not propagated to the merit.

sample per-event luminosity (PEL)
Signal (2.223 +/- 0.133)E-6 fb-1
Background 1.773 E-10 fb-1
Samples from two LB were used, their inst. luminosity was averaged for the normalization.

The quantities in the merit function (S/sqrt(S+B)) can then be expressed:

S = (nPass_s / nTot_s)*(1 / PEL_s) = ε_s σ_s

B = (nPass_b / nTot_b)*(1 / PEL_b) = ε_b σ_b

A total of 2000 signal and 2000 background events were used for the optimization (nTot_s, nTot_b). Empty background events (containing no tracks) are subtracted in the efficiency denominator in B.

The cuts optimized in this run were:

Name Range - low Range - high n steps
track_pT_min 3500 6000 50
track_chi2_max 0 20 40
minAbsD0_min 0 1.2 120
ZCA_max 0 0.6 60
invMassLo_min 4000 4700 35
invMassHi_max 5600 7000 70
pT_min 5000 15000 50
alpha2D_max 0 1 100
lxy_min -2 8 50

Since lxy was used, the decay length was not optimized. It still being cut on, but with a fixed cut of -1000 that allows for everything to pass (easier than implementing it differently). Implicitly, the requirement on B candidate neutrality is applied. 5 optimizing iterations were performed, and the resulting merit scans and cutflow is shown in the following figures.

visIter graph preliminary.png
Cut-value scans of merit function in 5 iterations
cutFlow preliminary.png
Cut flow for signal and background EVENTS, no number means 0
Lxy "mystery": the flat behavior below zero is legitimate: all sub-zero contributions to S are removed by the pointing angle cut, and the BKG events have at least one candidate with positive Lxy.

Further optimization runs

Variable distributions with 2 GeV track pT cut and 50 Chi2/ndf

trigVars-2GeVTrkCut pdf-likeScaling large.png
PDF-like scaling - histograms normalized by number of entries in all bins (including under/over flow)
trigVars-2GeVTrkCut trueScaling large.png
True scaling - histograms normalized to 1 fb-1 using per-event luminosities and numbers of events, background is further downscaled by 1E6 in track variables and 1E8 in candidate variables.

PDF-like plots for track pT cut 2, 3, 4 GeV

Alternative SV definition - projection of midpoint of the line connecting closest-approach points on the two tracks.

Two optimization runs were performed, but they ran out of background events in the optimization process.

visIter graph oldSV 10Iter.png
merit plots
cutFlow oldSV 10Iter.png
cut flow
notes: added abs(abs(d01) - abs(d02)) variable, larger range on track pT
visIter graph redefinedSV 8Iter.png
merit plots
cutFlow redefinedSV 8Iter.png
cut flow
notes: added abs(abs(d01) - abs(d02)) variable, larger range on track pT, SV redefined - midpoint of closest approach

<a name="lookHere"></a>CORRECTED BKG XSection: Run with proper 0-th iteration & larger BKG statistics

UPDATE: bkg normalization was fixed - the rate in our EB sample was about 330 Hz instead of 100 kHz used in the section above. More Bkg events were used from different LB, and the B-term in the merit function was calculated as weighted average of B-terms of individual LB, weight is the total number of events used from each LB (currently using same number of events in each LB).

LB # nTot lumi [fb-1s-1] rate
130 1000 1.761e-5 339
137 1000 1.729e-5 355
151 1000 1.786e-5 340
178 1000 1.711e-5 332
B-termCalc.jpg
B-term calculation details

In the following plots, the discriminating variables are plotted. Histograms are normalized by the integrated luminosity of the sample.

varPlots logScale 3-5GeVtrpTCut fix.png
Discriminating variable distribution, no bkg downscale, log scale

Iteration 0 had been performed - scanning the merit with loose cuts on the discriminating variables. Those were 3.5 GeV on track pT, max 50 track chi2, B-neutrality, invariant mass in range 4000 MeV - 7000 MeV. Other cuts were set to values that are extremely loose to prevent them from interfering, namely: ZCA_max = 20 mm, lxy_min = -1000, d0AbsDiff_max = 50 mm. The MC-generation cuts are listed in the MC JO: https://gitlab.cern.ch/atlas-physics/pmg/infrastructure/mc15joboptions/blob/master/share/DSID300xxx/MC15.300760.Pythia8B_A14_CTEQ6L1_B_hh.py: pTHatMin = 5 Gev,

(X, Y, Z) = (rangeLo, rangeHi, nStep) Cuts for 0th iteration Cuts for 1st iteration Cuts for 2nd iteration Cuts for 3rd iteration
track pT min 3.5 GeV (3.5, 6, 50) 3.5 GeV (3.5, 6, 50) 3.5 GeV (3.5, 6, 50) 3.6 GeV (3.5, 6, 50)
track chi2 max 50 (0, 20, 100) 50 (0, 20, 100) 2.2 (0, 20, 100) 2.2 (0, 10, 100)
B-neutrality        
invMass min 4 GeV 4 GeV 4 GeV 4 GeV
invMass max 7 GeV 7 GeV 7GeV 7 GeV
ZCA max 20 mm (0, 0.6, 60) 0.15 mm (0, 0.6, 60) 0.15 mm (0, 0.6, 60) 0.09 mm (0, 0.4, 80)
minAbsd0 min 0 mm (0, 1, 50) 0 mm (0, 1, 50) 0 mm (0, 1, 100) 0.06 mm (0, 0.4, 80)
d0AbsDiff max 50 mm (0, 1.8, 90) 50 mm (0, 1.8, 90) 0.42 mm (0, 1.8, 90) 0.6 mm (0, 1.8, 90)
lxy min -1000 mm (-2, 5, 60) -1000 mm (-2, 5, 60) -1000 mm (-2, 5, 60) 0.1 mm (-0.5, 1.5, 100)
a2d max pi (0, pi, 100) pi (0, pi, 100) pi (0, pi, 100) 0.0942478 (0, 1, 50)
cand pT 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 6.4 GeV (4, 15, 55)
MC generation cuts
pTHat min 5 GeV      
antiquark pt min 7 GeV      
antiquark eta max 2.6      
final hadrons pT 3.5 GeV      
final hadrons eta max 2.6      
Efficiency vs. lumi values: fitVals effVsInstLumi_it1cuts_fit.png effVsInstLumi_it2cuts_fit.png effVsInstLumi_it3cuts_fit.png

These are the scans of the merit function:

Iteration 0
Iteration 0 with loose cuts - merit plots with effS & effB overlaid Iteration 0 with loose cuts - merit plots with S and B cross-sections overlaid, B is scaled down by 10

These plots only suggest a cut on ZCA at 0.15 mm, no other maxima are present. In the next iteration therefore this ZCA cut was introduced:

Iteration "1"  
Iteration 1 with 0.15 ZCA cut - merit plots with effS & effB overlaid Iteration 1 with 0.15 ZCA cut - merit plots with S and B cross-sections overlaid, B is scaled down by 10 resulting cutFlow

Other merit maxima appeared, such as in chi2 and absD0Diff. Increasing tail disappeared from the high values of pointing angle. The pronounced maxima were used for the next iteration, the ranges were still kept at the same values.

Iteration "2"  
Iteration 2 merit plots with effS & effB overlaid Iteration 2 merit plots with S and B cross-sections overlaid, B is scaled down by 10 resulting cutFlow

For the next iteration, ranges of certain variables were adjusted and some scans were performed in finer steps.

Iteration "3"  
Iteration 3 merit plots with effS & effB overlaid Iteration 3 merit plots with S and B cross-sections overlaid, B is scaled down by 10 resulting cutFlow
Notably, the candidate pT and lxy peaks are not there anymore.

Optimization with full bkg sample

For the next runs, the following merit function is used:

merit = S / sqrt( S + B )

S = (nPass_s / nTot_s)*(1 / PEL_s) = ε_s σ_s * l_inst * 1/collRate corresponding to number of signal events to be accepted per second

B = (nPass_b / nTot_b)*(1 / PEL_b) = ε_b corresponding to number of bkg events accepted per second

To match their descriptions, these values should be multiplied by the L1 rate that our selection will run on, but that's common for both terms and thus irrelevant for the merit maxima.

This merit function did not change the cut values in the iterations 1 and 2 presented in the previous section. Iteration 3 was already running out of bkg events.

The ful EB360026 sample of 1.1M events was employed for the following section. The tracking of the full EB360026 sample was performed by athena release 22.0.25 as opposed to previous section, where this was 22.0.18. There are differences in terms of tracking which are being investigated in parallel, but the bottomline is that release 22.0.25 is newer and therefore more suitable for further studies. The most immediate effect is, that release 22.0.18 produced more tracks in the fullscans compared to 22.0.25.

The signal sample is still produced with release 21.3, and in this section 100k events was used.

numbers of events for run 3 are calculated as follows:

N_s = ε_s * (1 / PEL_s) * L_int * (L1 rate / collRate)

N_b = ε_b * L1 rate * (L_int / l_inst)

where L_int = 200 fb-1, l_inst = 2 E34 cm-2 s-1, L1 rate = 100 kHz, collRate = 30 MHz

(X, Y, Z) = (rangeLo, rangeHi, nStep) Starting Cuts for 0th iteration Starting Cuts for 1st iteration Starting Cuts for 2nd iteration Starting Cuts for 3rd iteration Starting Cuts for 4th iteration Starting Cuts for 5th iteration Starting Cuts for 6th iteration
track pT min 3.5 GeV (3.5, 6, 50) 3.5 GeV (3.5, 6, 50) 3.65 GeV (3.5, 6, 50) 3.8 GeV (3.5, 6, 50) 3.75 GeV (3.5, 6, 50) 3.95 GeV (3.5, 6, 50) 3.95 GeV (3.5, 6, 50)
track chi2 max 50 (0, 20, 100) 50 (0, 20, 100) 2.2 (0, 10, 50) 1.4 (0, 5, 50) 1.4 (0, 5, 50) 1.4 (0, 5, 50) 1.4 (0, 5, 50)
B-neutrality              
invMass min 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV
invMass max 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV
ZCA max 20 mm (0, 0.6, 60) 0.11 mm (0, 0.6, 60) 0.09 mm (0, 0.3, 30) 0.07 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30)
minAbsd0 min 0 mm (0, 1, 50) 0 mm (0, 1, 50) 0.06 mm (0, .5, 50) 0.08 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30)
d0AbsDiff max 50 mm (0, 1.8, 90) 50 mm (0, 1.8, 90) 0.52 mm (0, 1, 100) 0.21 mm (0, 0.8, 80) 0.22 mm (0, 0.8, 80) 0.21 mm (0, 0.8, 80) 0.21 mm (0, 0.8, 80)
lxy min -1000 mm (-2, 5, 60) -1000 mm (-2, 5, 60) .1 mm (-.5, 1, 150) .17 mm (-.1, .4, 50) .17 mm (-.1, .4, 50) .17 mm (-.1, .4, 50) .17 mm (-.1, .4, 50)
a2d max pi (0, pi, 100) 0.0942478 (0, 1, 100) .17 (0, 0.5, 50) 0.08 (0, 0.3, 30) 0.09 (0, 0.3, 30) 0.1 (0, 0.3, 30) 0.09 (0, 0.3, 30)
cand pT 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55)
MC generation cuts
pTHat min 5 GeV            
antiquark pt min 7 GeV            
antiquark eta max 2.6            
final hadrons pT 3.5 GeV            
final hadrons eta max 2.6            
Signal efficiency 78.998 % 58.826 % 37.451 % 17.965 % 16.801 % 14.944 % 14.803 %
Background rejection (no lumi extrapolation) 19.2511 % 82.6746 % 99.367279 % 99.9273305 % 99.939427 % 99.9526971 % 99.9535998 %
Number of signal events (run3) 237231 +/- 14218 176655 +/- 10594 112465 +/- 6753 53948.9 +/- 3253 50453.5 +/- 3043 44876.9 +/- 2709.8 44453.5 +/- 2684.5
Number of bkg events (lumi extrapolation for run3) 8.91784e+11 +/- 3.22369e+10 2.42267e+11 +/- 2.65561e+10 1.06739e+10 +/- 1.80477e+09 1.30056e+09 +/- 3.14244e+08 1.20518e+09 +/- 3.11497e+08 9.57643e+08 +/-2.55639e+08 9.45873e+08 +/-2.52804e+08
Efficiency vs. lumi   Iteration 1 starting cuts - efficiency vs. inst. lumi Iteration 2 starting cuts - efficiency vs. inst. lumi Iteration 3 starting cuts - efficiency vs. inst. lumi Iteration 4 starting cuts - efficiency vs. inst. lumi Iteration 5 starting cuts - efficiency vs. inst. lumi Iteration 6 starting cuts - efficiency vs. inst. lumi
Merit scans (B scaled down by 10) Iteration 0 merit plots with S & B overlaid Iteration 1 merit plots with S & B overlaid Iteration 2 merit plots with S & B overlaid Iteration 3 merit plots with S & B overlaid Iteration 4 merit plots with S & B overlaid Iteration 5 merit plots with S & B overlaid Iteration 6 merit plots with S & B overlaid
Merit scans (efficiency overlaid) Iteration 0 merit plots with effS & effB overlaid Iteration 1 merit plots with effS & effB overlaid Iteration 2 merit plots with effS & effB overlaid Iteration 3 merit plots with effS & effB overlaid Iteration 4 merit plots with effS & effB overlaid Iteration 5 merit plots with effS & effB overlaid Iteration 6 merit plots with effS & effB overlaid
Cut flow Iteration 0 - resulting cut flow Iteration 1 - resulting cut flow Iteration 2 - resulting cut flow Iteration 3 - resulting cut flow Iteration 4 - resulting cut flow Iteration 5 - resulting cut flow Iteration 6 - resulting cut flow
at 100 kHz L1 rate, this selection would fire at 94 +/- 25 Hz at inst lumi 2E34 cm-2 s-1

In the following table, optimization without track pT and pointing angle is performed to see if the cand pT and lxy - potentially highly correlated with track pT and a2d respectively - lead to better performance.

(X, Y, Z) = (rangeLo, rangeHi, nStep) Starting Cuts for 0th iteration Starting Cuts for 1st iteration Starting Cuts for 2nd iteration Starting Cuts for 3rd iteration Starting Cuts for 4th iteration Starting Cuts for 5th iteration Starting Cuts for 6th iteration
track pT min 3.5 GeV 3.5 GeV 3.5 GeV 3.5 GeV 3.5 GeV 3.5 GeV 3.5 GeV
track chi2 max 50 (0, 20, 100) 50 (0, 20, 100) 50 (0, 10, 50) 1.6 (0, 5, 50) 1.5 (0, 5, 50) 1.5 (0, 5, 50) 1.5 (0, 5, 50)
B-neutrality              
invMass min 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV 4 GeV
invMass max 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV 7 GeV
ZCA max 20 mm (0, 0.6, 60) 0.11 mm (0, 0.6, 60) 0.11 mm (0, 0.3, 30) 0.06 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30) 0.05 mm (0, 0.3, 30)
minAbsd0 min 0 mm (0, 1, 50) 0 mm (0, 1, 50) 0.06 mm (0, .5, 50) 0.06 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30) 0.09 mm (0, 0.3, 30)
d0AbsDiff max 50 mm (0, 1.8, 90) 50 mm (0, 1.8, 90) 50 mm (0, 1, 100) 0.24 mm (0, 0.8, 80) 0.28 mm (0, 0.8, 80) 0.22 mm (0, 0.8, 80) 0.22 mm (0, 0.8, 80)
lxy min -1000 mm (-2, 5, 60) -1000 mm (-2, 5, 60) .1 mm (-.5, 1.5, 150)

0.126667

mm (-.1, .5, 60)
.22 mm (-.1, .5, 60) .2 mm (-.1, .5, 60) .22 mm (-.1, .5, 60)
a2d max pi pi pi pi pi pi pi
cand pT 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55) 6100 GeV (4, 15, 55) 0 GeV (4, 15, 55) 0 GeV (4, 15, 55)
MC generation cuts
pTHat min 5 GeV            
antiquark pt min 7 GeV            
antiquark eta max 2.6            
final hadrons pT 3.5 GeV            
final hadrons eta max 2.6            
Signal efficiency             20.228 %
Background rejection (no lumi extrapolation)             99.75 %
Number of signal events (run3)             60744.7 +/- 3659.15
Number of bkg events (lumi extrapolation for run3)             2.50806e+09 +/-4.99195e+08
Efficiency vs. lumi             Iteration 6 starting cuts - efficiency vs. inst. lumi
Merit scans (B scaled down by 10)   Iteration 1 merit plots with S & B overlaid Iteration 2 merit plots with S & B overlaid Iteration 3 merit plots with S & B overlaid Iteration 4 merit plots with S & B overlaid Iteration 5 merit plots with S & B overlaid Iteration 6 merit plots with S & B overlaid
Merit scans (efficiency overlaid)   Iteration 1 merit plots with effS & effB overlaid Iteration 2 merit plots with effS & effB overlaid Iteration 3 merit plots with effS & effB overlaid Iteration 4 merit plots with effS & effB overlaid Iteration 5 merit plots with effS & effB overlaid Iteration 6 merit plots with effS & effB overlaid
Cut flow   Iteration 1 - resulting cut flow Iteration 2 - resulting cut flow Iteration 3 - resulting cut flow Iteration 4 - resulting cut flow Iteration 5 - resulting cut flow Iteration 6 - resulting cut flow
at 100 kHz L1 rate, this selection would fire at 251 +/- 49 Hz at inst lumi 2E34 cm-2 s-1

Additional variables

The baseline selection is that of

tr pT > 3950 GeV mm
tr chi2/dof < 1.4 mm
minAbsd0 > 0.09 mm
ZCA < 0.05 mm
mass 4 GeV < < 7 GeV
a2d < 0.1
d0 asym < 0.21 mm
It was obtained in the optimization process without cand_pT and lxy variables using signal sample tracked in release 22.0.24 rather than 21.3 as was the case in previous section. This has 15.356 % signal efficiency and 0.0474 % BKG efficiency (without luminosity extrapolation). This translates to 527 events passing the above selection.

Additional selection can be introduced to further limit the background. This includes the number of hits in pixel/SCT layers. With a nPix > 0 && nSCT > 1 (coincedes with Bmumu preselection) requirement, the background yield drops to 352 candidates at 14.939 % signal efficiency. These numbers correspond to no further optimization, only a simple cut on top of the existing baseline optimized cuts. When optimization is run with this additional requirement, the result oscillates between state 3900 GeV tr pT && 0.22 mm d0 asym and 3950 GeV tr pT && 0.16 mm d0 asym.


-- OndrejKovanda - 2020-05-20

Topic attachments
I Attachment History Action Size Date Who Comment
JPEGjpg 20201006_180428.jpg r1 manage 1207.1 K 2020-10-06 - 18:08 OndrejKovanda  
PNGpng B-peak_4GeVTr_7GeVB_1mmDZ_rec.png r1 manage 6.9 K 2020-07-15 - 15:51 OndrejKovanda  
PNGpng B-peak_4GeVTr_7GeVB_1mmDZ_ref.png r1 manage 7.1 K 2020-07-15 - 15:52 OndrejKovanda  
JPEGjpg B-termCalc.jpg r1 manage 3457.4 K 2021-01-18 - 14:48 OndrejKovanda  
PNGpng BCandDZ_truth.png r1 manage 9.9 K 2020-09-08 - 17:52 OndrejKovanda  
PNGpng BCandLT_correctTruth.png r1 manage 12.1 K 2020-09-10 - 16:12 OndrejKovanda  
PNGpng BCandLT_matchCheck.png r1 manage 13.4 K 2020-09-16 - 16:29 OndrejKovanda matchCheck
PNGpng BCandLT_truth.png r1 manage 12.2 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng BCandMass_truth.png r1 manage 11.9 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng BCandMass_truth2.png r1 manage 11.7 K 2020-09-08 - 17:51 OndrejKovanda  
PNGpng BCandMass_weird.png r1 manage 12.1 K 2020-09-07 - 14:33 OndrejKovanda  
PNGpng BCandPA_correctTruth.png r1 manage 11.9 K 2020-09-10 - 16:12 OndrejKovanda  
PNGpng BCandPA_matchCheck.png r1 manage 11.6 K 2020-09-16 - 16:29 OndrejKovanda matchCheck
PNGpng BCandPA_truth.png r1 manage 13.5 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng BCand_DZ0_4GeV_with_Bs_noCut.png r1 manage 13.2 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng BCand_DZ0_hi_range_4GeV_with_Bs_noCut.png r1 manage 14.8 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng BCand_d01_-_d02_4GeV_with_Bs_noCut.png r1 manage 16.2 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng BCandlxy_correctTruth.png r1 manage 11.1 K 2020-09-10 - 16:12 OndrejKovanda  
PNGpng BCandlxy_matchCheck.png r1 manage 12.6 K 2020-09-16 - 16:29 OndrejKovanda matchCheck
PNGpng BCandlxy_truth.png r1 manage 10.7 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng BCandpT_truth.png r1 manage 10.9 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng B_Candidate_DZ_at_closest_approach_4GeV_with_Bs_noCut.png r1 manage 16.9 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_Lxy_4GeV_with_Bs_noCut.png r1 manage 15.8 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_Pointing_Angle_4GeV_with_Bs_noCut.png r1 manage 15.0 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_invariant_mass_4GeV_with_Bs_noCut.png r1 manage 16.2 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_min_d0_4GeV_with_Bs_noCut.png r1 manage 16.2 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_pT_4GeV_with_Bs_noCut.png r1 manage 15.3 K 2020-11-23 - 09:25 OndrejKovanda  
PNGpng B_Candidate_propper_decay_length_4GeV_with_Bs_noCut.png r1 manage 16.0 K 2020-11-23 - 09:25 OndrejKovanda  
Unknown file formatxlsx Bhh.xlsx r1 manage 16.7 K 2021-01-06 - 12:47 OndrejKovanda  
PDFpdf FS_update-30-9-20-v2.pdf r1 manage 864.8 K 2020-10-06 - 17:20 OndrejKovanda comparisonPresentation
Unix shell scriptsh HITS2RDO.sh r1 manage 2.2 K 2020-05-20 - 13:48 OndrejKovanda  
PNGpng SVPosition_correctTruth.png r1 manage 11.0 K 2020-09-10 - 16:08 OndrejKovanda  
PNGpng SVPosition_truth.png r1 manage 11.1 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng Track_chi2-nDOF_4GeV_with_Bs_noCut.png r1 manage 13.5 K 2020-11-23 - 09:26 OndrejKovanda  
PNGpng Track_chi2-nDOF_noCut.png r1 manage 13.6 K 2020-10-29 - 09:28 OndrejKovanda  
PNGpng Track_d0_4GeV_with_Bs_noCut.png r1 manage 14.3 K 2020-11-23 - 09:26 OndrejKovanda  
PNGpng Track_d0_noCut.png r1 manage 14.4 K 2020-10-29 - 09:28 OndrejKovanda  
PNGpng Track_pT_4GeV_with_Bs_noCut.png r1 manage 15.6 K 2020-11-23 - 09:26 OndrejKovanda  
PNGpng Track_pT_noCut.png r1 manage 15.4 K 2020-10-29 - 09:28 OndrejKovanda  
Texttxt cutFlow.txt r1 manage 2.3 K 2020-11-16 - 15:50 OndrejKovanda  
PNGpng cutFlow_oldSV_10Iter.png r1 manage 71.6 K 2021-01-12 - 10:16 OndrejKovanda  
PNGpng cutFlow_preliminary.png r1 manage 47.4 K 2021-01-06 - 15:49 OndrejKovanda  
PNGpng cutFlow_redefinedSV_8Iter.png r1 manage 62.4 K 2021-01-12 - 10:17 OndrejKovanda  
PNGpng cutFlow_run1_10Iter.png r1 manage 76.0 K 2021-01-14 - 10:47 OndrejKovanda  
PNGpng cutFlow_run1_Iterations5-6-7.png r1 manage 38.4 K 2021-01-14 - 10:50 OndrejKovanda  
PNGpng cutFlow_run2_iteration0-5_ZCA0-15.png r1 manage 30.6 K 2021-01-18 - 16:17 OndrejKovanda  
PNGpng cutFlow_run2_iteration2.png r1 manage 30.2 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng cutFlow_run2_iteration3.png r1 manage 29.4 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng cutFlow_run2_iteration_0_fixed.png r1 manage 31.6 K 2021-01-26 - 09:56 OndrejKovanda  
PNGpng cutFlow_run2_iteration_1_fixed.png r1 manage 30.7 K 2021-01-26 - 09:56 OndrejKovanda  
PNGpng cutFlow_run2_iteration_2_fixed.png r1 manage 29.9 K 2021-01-26 - 09:56 OndrejKovanda  
PNGpng cutFlow_run2_iteration_3_fixed.png r1 manage 29.4 K 2021-01-26 - 09:56 OndrejKovanda  
PNGpng cutFlow_run4_iteration_0_superHiStat.png r2 r1 manage 29.8 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_1_superHiStat.png r1 manage 28.3 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_2_superHiStat.png r1 manage 29.6 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_3_superHiStat.png r1 manage 27.6 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_4_superHiStat.png r1 manage 29.2 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_5_superHiStat.png r1 manage 29.0 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run4_iteration_6_superHiStat.png r1 manage 29.0 K 2021-02-11 - 10:22 OndrejKovanda  
PNGpng cutFlow_run5_iteration_1_candPt_lxy.png r1 manage 24.4 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutFlow_run5_iteration_2_candPt_lxy.png r1 manage 24.3 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutFlow_run5_iteration_3_candPt_lxy.png r1 manage 24.5 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutFlow_run5_iteration_4_candPt_lxy.png r1 manage 24.1 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutFlow_run5_iteration_5_candPt_lxy.png r1 manage 24.1 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutFlow_run5_iteration_6_candPt_lxy.png r1 manage 24.1 K 2021-02-16 - 10:25 OndrejKovanda  
PNGpng cutsCompare-matched.png r1 manage 9.5 K 2020-07-08 - 13:05 OndrejKovanda  
PNGpng cutsCompare.png r1 manage 9.8 K 2020-07-08 - 13:05 OndrejKovanda  
PNGpng effEvol_iteration_1_cuts.png r1 manage 16.4 K 2021-02-11 - 10:45 OndrejKovanda  
PNGpng effEvol_iteration_2_cuts.png r1 manage 18.6 K 2021-02-11 - 10:45 OndrejKovanda  
PNGpng effEvol_iteration_3_cuts.png r1 manage 18.1 K 2021-02-11 - 10:45 OndrejKovanda  
PNGpng effEvol_iteration_4_cuts.png r1 manage 17.4 K 2021-02-11 - 10:45 OndrejKovanda  
PNGpng effEvol_iteration_5_cuts.png r1 manage 15.0 K 2021-02-16 - 10:01 OndrejKovanda  
PNGpng effEvol_iteration_6_cuts.png r1 manage 16.8 K 2021-02-16 - 10:01 OndrejKovanda  
PNGpng effEvol_run4_it1.png r1 manage 14.7 K 2021-02-09 - 10:51 OndrejKovanda  
PNGpng effVsInstLumi_it1cuts_fit.png r1 manage 15.6 K 2021-02-09 - 09:43 OndrejKovanda  
PNGpng effVsInstLumi_it2cuts_fit.png r1 manage 16.5 K 2021-02-09 - 09:43 OndrejKovanda  
PNGpng effVsInstLumi_it3cuts_fit.png r1 manage 16.6 K 2021-02-09 - 09:43 OndrejKovanda  
Unknown file formatext fitVals r1 manage 0.7 K 2021-02-09 - 09:51 OndrejKovanda  
PNGpng histos300431.png r1 manage 37.7 K 2020-11-30 - 08:59 OndrejKovanda  
PNGpng histos300760.png r1 manage 36.5 K 2020-11-30 - 08:59 OndrejKovanda  
PNGpng histos_pTHat_300431.png r1 manage 34.5 K 2020-12-07 - 16:21 OndrejKovanda  
PNGpng histos_pTHat_300760.png r1 manage 34.2 K 2020-12-07 - 16:21 OndrejKovanda  
PNGpng invMass2GeVTr_2GeVB_rec.png r1 manage 8.9 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass2GeVTr_2GeVB_ref.png r1 manage 9.1 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass4GeVTr_4GeVB_rec.png r1 manage 8.2 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass4GeVTr_4GeVB_ref.png r1 manage 8.7 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass4GeVTr_7GeVB_1mmDZ_rec.png r1 manage 7.6 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass4GeVTr_7GeVB_1mmDZ_ref.png r1 manage 7.9 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass_noCuts_rec.png r1 manage 8.2 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng invMass_noCuts_ref.png r1 manage 8.0 K 2020-07-14 - 11:40 OndrejKovanda cuts
PNGpng run5_effEvol_iteration_6_cuts.png r1 manage 15.8 K 2021-02-16 - 10:39 OndrejKovanda  
PNGpng scatter_mind0vsDZ0_chi2-dof_l5.png r2 r1 manage 39.9 K 2020-11-23 - 09:26 OndrejKovanda  
PNGpng trackD0_truth.png r1 manage 10.9 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng trackpT_truth.png r1 manage 11.4 K 2020-09-08 - 17:46 OndrejKovanda  
PNGpng trigVars-2GeVTrkCut_pdf-likeScaling.png r1 manage 95.7 K 2021-01-12 - 10:07 OndrejKovanda  
PNGpng trigVars-2GeVTrkCut_pdf-likeScaling_large.png r1 manage 139.1 K 2021-01-12 - 09:59 OndrejKovanda  
PNGpng trigVars-2GeVTrkCut_trueScaling_large.png r1 manage 143.3 K 2021-01-12 - 09:59 OndrejKovanda  
PNGpng trigVars-3GeVTrkCut_pdf-likeScaling.png r1 manage 96.1 K 2021-01-12 - 10:07 OndrejKovanda  
PNGpng trigVars-4GeVTrkCut_pdf-likeScaling.png r1 manage 95.2 K 2021-01-12 - 10:07 OndrejKovanda  
PNGpng varPlots_logScale_3-5GeVtrpTCut_fix.png r1 manage 90.9 K 2021-01-18 - 15:25 OndrejKovanda  
PNGpng visIter_graph.png r1 manage 128.7 K 2020-11-16 - 15:50 OndrejKovanda  
PNGpng visIter_graph_SB_iteration0.png r1 manage 112.1 K 2021-01-15 - 15:32 OndrejKovanda  
PNGpng visIter_graph_SB_run2_iteration0-5_ZCA0-15.png r1 manage 93.3 K 2021-01-18 - 16:09 OndrejKovanda iter0-5
PNGpng visIter_graph_SB_run2_iteration0_finer.png r1 manage 123.4 K 2021-01-18 - 14:55 OndrejKovanda  
PNGpng visIter_graph_SB_run2_iteration2.png r1 manage 99.7 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng visIter_graph_SB_run2_iteration3.png r1 manage 102.8 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng visIter_graph_SB_run2_iteration_0_fixed.png r1 manage 93.6 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_SB_run2_iteration_1_fixed.png r1 manage 93.0 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_SB_run2_iteration_2_fixed.png r1 manage 101.9 K 2021-01-26 - 10:00 OndrejKovanda  
PNGpng visIter_graph_SB_run2_iteration_3_fixed.png r1 manage 106.2 K 2021-01-26 - 10:00 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_0_superHiStat.png r2 r1 manage 105.5 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_1_superHiStat.png r1 manage 106.2 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_2_superHiStat.png r1 manage 98.6 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_3_superHiStat.png r1 manage 95.5 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_4_superHiStat.png r1 manage 95.3 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_5_superHiStat.png r1 manage 95.8 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run4_iteration_6_superHiStat.png r1 manage 95.7 K 2021-02-11 - 10:24 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_1_candPt_lxy.png r1 manage 82.1 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_2_candPt_lxy.png r1 manage 83.2 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_3_candPt_lxy.png r1 manage 73.9 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_4_candPt_lxy.png r1 manage 73.2 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_5_candPt_lxy.png r1 manage 74.3 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_SB_run5_iteration_6_candPt_lxy.png r1 manage 74.2 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run1_iteration0.png r1 manage 116.1 K 2021-01-15 - 15:32 OndrejKovanda  
PNGpng visIter_graph_eff_run2_iteration0-5_ZCA0-15.png r1 manage 130.6 K 2021-01-18 - 16:09 OndrejKovanda iter0-5
PNGpng visIter_graph_eff_run2_iteration0_finer.png r1 manage 128.1 K 2021-01-18 - 14:55 OndrejKovanda  
PNGpng visIter_graph_eff_run2_iteration2.png r1 manage 129.2 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng visIter_graph_eff_run2_iteration3.png r1 manage 124.9 K 2021-01-21 - 09:20 OndrejKovanda  
PNGpng visIter_graph_eff_run2_iteration_0_fixed.png r1 manage 128.0 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_eff_run2_iteration_1_fixed.png r1 manage 129.4 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_eff_run2_iteration_2_fixed.png r1 manage 130.6 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_eff_run2_iteration_3_fixed.png r1 manage 128.0 K 2021-01-26 - 09:55 OndrejKovanda visIters
PNGpng visIter_graph_eff_run4_iteration_0_superHiStat.png r2 r1 manage 126.7 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_1_superHiStat.png r1 manage 121.5 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_2_superHiStat.png r1 manage 114.8 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_3_superHiStat.png r1 manage 110.5 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_4_superHiStat.png r1 manage 110.2 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_5_superHiStat.png r1 manage 111.1 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run4_iteration_6_superHiStat.png r1 manage 111.3 K 2021-02-11 - 10:23 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_1_candPt_lxy.png r1 manage 94.6 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_2_candPt_lxy.png r1 manage 95.2 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_3_candPt_lxy.png r1 manage 85.3 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_4_candPt_lxy.png r1 manage 84.5 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_5_candPt_lxy.png r1 manage 84.8 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_eff_run5_iteration_6_candPt_lxy.png r1 manage 84.6 K 2021-02-16 - 10:26 OndrejKovanda  
PNGpng visIter_graph_efficiency.png r1 manage 100.9 K 2021-01-15 - 11:10 OndrejKovanda  
PNGpng visIter_graph_iter0_base.png r1 manage 58.3 K 2021-01-14 - 10:38 OndrejKovanda  
PNGpng visIter_graph_oldSV_10Iter.png r1 manage 418.6 K 2021-01-12 - 10:14 OndrejKovanda  
PNGpng visIter_graph_preliminary.png r1 manage 245.7 K 2021-01-06 - 15:50 OndrejKovanda  
PNGpng visIter_graph_redefinedSV_8Iter.png r1 manage 332.2 K 2021-01-12 - 10:15 OndrejKovanda  
PNGpng visIter_graph_run1_10Iter.png r1 manage 420.4 K 2021-01-14 - 10:47 OndrejKovanda  
PNGpng visIter_graph_run1_Iteration5-6-7.png r1 manage 142.7 K 2021-01-14 - 10:50 OndrejKovanda  
Edit | Attach | Watch | Print version | History: r46 < r45 < r44 < r43 < r42 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r46 - 2021-03-05 - OndrejKovanda
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback