Particle Flow Workshop in Paris: 18-19 Oct 2007

Complete: 4

Venue

Paris (France), 18-19 October 2007

Agenda

Agenda, registration form, list of hotels, ... can be found here (under construction)

Plan of work for the WORKshop (and beyond)

Preshower clustering

In the endcaps, electrons and photons often leave significant signals in at least one - and often both - layers of the preshower. The strips must therefore be clusterized and linked to charged particle tracks in a way similar to what is done with ECAL or HCAL clusters. Preshower clusters must also be linked with ECAL clusters. These preshower signals are useful to

  1. Identify electrons (with respect to charged hadrons, which are not expected to leave signals there)
  2. Identify photons merged in charged hadron ECAL clusters
  3. Determine the energy of photons and electrons
  4. Improve the direction of photons (especially for Bremsstrahlung photons when it comes to link them with the parent electron)

A preliminary work was already done with FAMOS a couple years ago, but this work must be ported, and most likely revisited, in CMSSW. A large part of the work can be developed with Fast Simulation, and eventually checked with Full Simulation.

Work to be done (and presented at the workshop):

  1. With single photons and single electrons, port/revisit the clustering algorithm developed with FAMOS. Produce the related calibration maps.
  2. With single photons, port/revisit the linking algorithm, between the two layers first, and with the ECAL clusters next. Produce the related resolution maps.
  3. With single electrons (first by switching off the material effects), port/revisit the linking algorithm with charged particle tracks. Produce related resolution maps.
  4. Develop an algorithm to determine the energy of electrons and photons.
  5. Develop an algorithm to determine their directions
  6. Include the above in particle flow algorithm and see the effects on taus and jets in the endcaps
  7. Check the effect in the electron (pre)Id algorithm

People involved (please volunteer!) : Noone, yet

Cluster energy calibration

The PFClusters must, as the CMS.CaloTowers, be calibrated. The reason is threefold.

  • Isolated neutral hadrons

First, isolated neutral hadrons (i.e., not linked to a charged particle track) deposit their energy in ECAL and/or HCAL. Since the ECAL is primarily calibrated for photons, and the HCAL is originally calibrated to give the same response as that of the ECAL to hadrons, the measured raw energy is not that of the orginal hadron, and a calibration function f has to be used,

Ehadron = f(EECAL, EHCAL).

Hadrons that leave energy only in the HCAL do so because they have not interacted in the ECAL before. As a consequence, the fraction of energy lost in the material between the ECAL and the HCAL tends to zero in this case, and a different calibration function is needed,

Ehadron = g(EHCAL).

These calibration functions have been determined two years ago with FAMOS (i.e., the COBRA/ ORCA ancestor of the Fast Simulation). A non-linear form was determined for the function g, while the function f was for the sake of simplicity assumed to be linear,

Ehadron = a(eta) + b(eta) EECAL + c(eta)EHCAL.

The three coefficients are expected to depend on pseudo-rapidity, and the b and c parameters to be very close to each other. This linear form appears to overestimate the energy of hadrons below 5 GeV, by a factor as large as 2 for a 1 GeV hadron. Since this could lead to an important bias of the MET measurement, a non linear form should be determined for f as well. Moreover, all parameters should be determined again using the fast and full simulation of CMSSW, and the exercise of determining all these coefficients and functions directly from the data should be worked on.

  • Merged neutral hadrons

Second, the energy of neutral hadrons merged in charged clusters has also to be determined. Basically, this energy is the difference between the calibrated ECAL and HCAL energies (calibrated as explained above) and the sum of the charged particle momenta linked to this (these) cluster(s). This difference might be, however, either negative or statistically insignificant, in which case no neutral hadron must be counted. The siginificance is therefore to be determined from the expected response spread to, e.g., isolated charged or neutral hadrons:

sigma(Ehadron) = s(EECAL,EHCAL,eta)

and a neutral hadron is created only if the above difference is larger than, e.g., 1, 2, or 3 sigmas. This cut is to be pragmatically determined to optimize the response to jets and MET

  • Merged photons

Photons merged in charged clusters (this happens often for very energetic taus/jets) must be treated and found before neutral hadrons are searched for, to avoid a spurious overcalibration of the former as if they were neutral hadrons. To identify these photons, the energy measured a ECAL PFCluster must first be compared to that expected to be deposited by all the charged particles linked to this cluster (under the hadron hypothesis), e.g.,

Eexpected = c(eta) + d(eta) ptracks

and the significance must be determined by comparing the difference between the expected and measured energy to the spread of the expected energy

sigma (Eexpected) = g(ptracks,eta)

If the difference is indeed significant (with a meaning to be pragmatically defined), a search for neutral hadron can be made as described in the previous bullet by comparing the sum of the calibrated HCAL and remaining ECAL energies (once the merged photons are removed) to the sum of the charged particle momenta linked to these deposits. If this difference is significantly negative, it may mean that, after all, there were no merged photons in this charged cluster. A multi-variate analysis involving the two differences and their significance might then give the final answer. (The inclusion of cluster shape variables, aimed at disentangling hadron and photon clusters, and angular chi squared, aimed at disentangling photons from charged clusters, might also serve the purpose.)

Work to be done (and presented at the workshop):

  • For neutral hadrons (in collaboration with HCAL DPG)

  1. Repeat the determination of the calibration functions previously done with FAMOS using Fast Simulation and single K0L. In particular, find a non linear form for the function f for low energy hadrons leaving energy both in ECAL and HCAL.
  2. Do the same with the full simulation, and compare the results
  3. Do the same with single charged pions, using the measured momentum for the true particle energy (probably with Fast Simulation if the previous two results do compare indeed). Compare the results with the above.
  4. Repeat the exercise with the minimum-bias events from the CSA07 production, with isolated charged particles instead of single charged pions from a particle gun. Compare the results, evaluate the statistical accuracy of the method, be ready for real data taking.
  5. Refine the existing algorithm that extracts neutral hadrons from charged clusters
  6. Check the effect on taus and jets

  • For merged photons (to be done in parallel with the above)

  1. Determine the maps of the ECAL expected energy and spread for single K0L, with fast and full simulation.
  2. Do the same with single charged pions and isolated charged particles in minimum-bias events and compare
  3. Develop an algorithm that extracts merged photons from charged clusters.
  4. Check the effect on taus and jets, especially at high ET

  • Charged momentum resolution

In all the above, the charged momentum is assumed to be infinitely good. For very large momenta and for tracks with a small number of hits, this assumption is no longer valid. The significance of the energy differences must therefore be determined by adding in quadrature the cluster energy resolution to the track momentum resolution.

People involved (please volunteer, there is work for many!) : Colin Bernet, Michel Della Negra, Alexandre Zabi

Track-cluster links

In the previous bullet, the link between charged particles and particle flow clusters is repeatedly mentioned. For the time being, a charged particle and a PFCluster are linked together if the chi squared made of the distance in eta and phi between the cluster position and the track extrapolated to the expected shower maximum (in the ECAL for ECAL clusters, in the HCAL for HCAL clusters) is smaller than a given value:

[etaCluster - etatrack]2 / sigma2(eta) + [phiCluster - phitrack]2 / sigma2(phi) < Value

where the sigma's and the cut Value depend on the track momentum, the energy of the cluster and most likely on the pseudo-rapidity too. Such resolution maps were made with FAMOS, and are implemented as such in the current particle-flow algorithm. Nothing was done with CMSSW (fast or full simulation) since. Tracks and clusters that share a track-cluster or cluster-cluster link are placed in a given "!PFBlock". Photons are defined as ECAL PFClusters with no direct link to a charged particle track. Isolated neutral hadrons are defined as HCAL PFClusters with no direct link to a charged particle track. (There may be some extra-calibration to be made for those photons linked to a isolated HCAL PFclusters, as they may come from the same neutral hadron, but this was never attempted.)

Charged and neutral hadrons (and soon, merged photons) are extracted from these PFBlocks as explained above. Because of early nuclear interaction (either just before or in the middle of the ECAL), the sum of the ECAL and HCAL energy linked to the original track(s) might much smaller than expected, and a number of satellite clusters might be wrongly identified as photons or neutral hadrons. For this reason, it is necessary to link these satellite clusters back to the charged particle track(s), until the linked (calibrated) calorimetric energy is compatible to the sum of the charged particle momenta, first in the HCAL (where the effect of early nuclear interactions are largest), second, and if needed in the ECAL. Another ordering could be according to increasing chi squared, as defined above.

Work to be done (and presented at the workshop):

  1. Repeat the resolution map building, previously done with FAMOS, with Fast Simulation amd with single K0L
  2. Do the same with the full simulation, and compare the results
  3. Do the same with single charged pions, using the measured momentum to determine the true particle position (probably with Fast Simulation if the previous two results do compare indeed). Compare the results with the above.
  4. Repeat the exercise with the minimum-bias events from the CSA07 production, with isolated charged particles instead of single charged pions from a particle gun. Compare the results, evaluate the statistical accuracy of the method, be ready for real data taking.
  5. Develop an algorithm that link satellites back to the original track
  6. Check the effect on taus and jets

People involved (please volunteer, there is work for many!) : Alexandre Zabi

Charged hadron energy

As already mentioned, the charged momentum is assumed to be infinitely good in all the above. It is so as well for the determination of the charged hadron energy, determined solely from the momentum of the reconstructed track. For very large momenta and for tracks with a small number of hits, this assumption is no longer valid. The charged hadron energy must therefore be computed as the weighted average of the track momentum and the calibrated cluster energies:

Ehadron = [ ptrack / sigma2(p) + EECAL+HCAL / sigma2(ECAL+HCAL) ] / [1/ sigma2(p) + 1/sigma2(ECAL+HCAL) ]

The charged hadron direction must be kept as that of the momentum determined at the main vertex.

Work to be done (and presented at the workshop):

  1. Implement the above in the algorithm
  2. Check the effect on taus and jets

People involved : Michel Della Negra

Nuclear interactions in the tracker

Nuclear interactions that occur just before or inside the ECAL lead to double counting that can be treated with the satellite recovery described above. Nuclear interactions that occur well inside the tracker give rise to a number of secondaries (charged and neutral), which would lead to severe double counting, not recoverable with the previous strategy, because the secondaries are well apart when they reach the calorimeters. It has to be kept in mind that about 20% of the hadrons (charged and neutral) do experience a nuclear interaction in the tracker. There will therefore be, in each jet, just below two such nuclear interactions on average! A new strategy has therefore to be thought of to minimize this double counting.

The simplest strategy would consist in identifying the nuclear interactions (stopping charged particle, followed by a high hit concentration), and not consider those tracks in the particle-flow reconstruction. (Note that this strategy is automatically applied for interacting neutral hadrons!) While indeed avoiding the aforementioned double counting, this strategy has two foreseeable drawbacks: (i) the double counting is replaced by missing energy, as nuclear interactions may give rise to neutrinos and slow particles that never reach calorimeters; (ii) the jet direction is degraded accordingly, and this effect is worsened by secondary charged particles either reconstructed as neutral (hence with the wrong direction) or not within the jet cone (hence missed for good).

The ideal strategy would be to reconstruct all secondary particles, and not consider them in the particle-flow reconstruction. Again, this strategy, while avoiding the double counting, has two foreseeable drawbacks: (i) Neutral secondary particles cannot be reconstructed (and not all the charged particles can be reconstructed), leading to residual double counting; (ii) Tracks with a small number of hits have a bad momentum determination, hence worsening the resolution.

It is only with a systematic and pragmatic study that what to be done in which conditions will be determined.

Work to be done (and presented at the workshop):

  1. Implement the nuclear interaction tagger in the particle flow sequence
  2. Implement the nuclear interaction reconstruction in the particle flow sequence
  3. Determine the timing of the above (might be prohibitive...)
  4. Develop an algorithm that uses the tagging information only and an algorithm that uses the reconstructed information
  5. Check the effect on taus and jets

People involved (more people needed!): Vincent Roberfroid

V0's and photon conversions

While work on V0 reconstruction with tracker-only seeding had started in the ORCA times (F.P. Schilling and A.S. Giolo-Nicollerat), this work has never been ported to CMSSW and certainly still needed a robust performance and timing optimization. This work should be carried out by the tracker DPG (tracking subgroup), but the motivation, and maybe part of the work, should come from us. The photon conversion reconstruction will benefit from this work as well.

Another approach is also followed for photon conversions (N. Marinelli), with ECAL seeding instead of tracker seeding. Whether this approach is suited for (low-energy) photon conversions in jets is currently being studied.

Work to be done:

  1. Develop a tracker-seeded V0 reconstruction algorithm (probably with the hits not used by the three-step tracking)
  2. Pursue the ECAL-seeded photon conversion algorithm development
  3. Implement in the particle-flow reconstruction
  4. See the effects on taus and jets

People involved : Nancy Marinelli (ECAL-seeded), ? (Tracker-seeded)

Electrons

An electron pre-identification, based on the three-step tracking and on basic ECAL and preshower information, is now available with 95% efficiency for electrons in jets and 5% efficiency for charged pions in jets. This is to be considered at the starting point for full electron reconstruction and identification in jets

Work to be done (and presented at the workshop):

  1. Identify the reason(s) for the remaining 5% of pions still pre-id'ed as electrons (nuclear interactions? charge exchange? ...)
  2. Possibly refine the pre-identification to take the above into account
  3. Apply the full GSF reconstruction to all electron candidates
  4. Port from FAMOS and continue to develop an algorithm that associate Bremsstrahlung photons to the electron tracks (compatible with directions tangent to the track and the energy lost at each tracker layer), first on single electrons, then on electrons in jets.
  5. Refine identification criteria by including the full GSF information, the number and energy of Brem photons, the cluster shape variables, the Eall/pin ratio (where Eall is the energy sum of the electron and all Brem photons, and pin is the track momentum at the main vertex), etc...
  6. Check the effect on taus and jets

Note: Most of the above can certainly be done with the Fast Simulation.

People involved: Michele Pioppi, Florian Beaudette

Muons

Muons are for now available in CMSSW under the form of global muons (i.e., muon hits and segments linked to tracker hits, all fit together to a single track), which could be used in the particle-flow reconstruction, at the very beginning of the algorithm:

  • The tracker hits used by these global muons must be removed from the hit sample used by the three-step tracking
  • The energy expected/measured in the ECAL crystal(s?) and the HCAL tower closest to the track extrapolation must be removed from the cells used in the ECAL and HCAL clustering algorithms.
  • A particle-flow candidate should be created (with PID +/- 13, mass 105.6 MeV/c2, and momentum given by the global fit) for later use in the AOD

Soon, so-called "tracker muons" will become fully available (a preliminary version already exist). These tracker muons start from tracker tracks (e.g., output of the three-step tracking), extrapolated first to the ECAL and the HCAL, then to the muon system, and a likelihood (wrt to the muon/pion hypothesis) is built from the ECAL crystal energies, the HCAL tower energies and the various muon hits and segments closest to the track extrapolation. These muons could then be used in the particle-flow reconstruction just after the three-step tracking, as follows:

  • A cut on the likelihood variable must be determined as a function of the muon pT and optimized with respect to the muon efficiency and pion purity (towards, e.g., the best MET resolution and tails)
  • The energy expected/measured in the ECAL crystal(s?) and the HCAL tower closest to the track extrapolation must be removed from the cells used in the ECAL and HCAL clustering algorithms.
  • A particle-flow candidate should be created (with PID +/- 13, mass 105.6 MeV/c2, and momentum given by the tracker fit) for later use in the AOD

It is even possible to use both algorithms in a row. Indeed, global muons are known to be rather pure, and are expected to have to best possible momentum determination, with all existing information. Once these gold-plated muons are found (and removed from tracker hits, ecal crystals and hcal towers), the remaining tracker muons could be looked for, towards an increase of the efficiency without loss of purity.

Work to be done (and presented at the workshop):

  1. Start with Global Muons and implement them in the particle-flow reconstruction as explained above.
  2. Check efficiency, purity and effect on taus, jets and MET (mostly tails)
  3. Then go for Tracker Muons with the three-step tracking tracks, and optimize the likelihood cut as explained above
  4. Check efficiency, purity and effects on the taus, jets and MET (mostly tails)

People involved (please volunteer, there is work for many!) : Martijn Mulders

Jet Clustering

The particle-flow algorithm produces individual and identified particles with much more information than is available from calorimeter cells alone and which have direct analogs to the visible, final-state, generator-level particles. The jet clustering algorithms should thus exploit this richness of information in the best possible way and in a way which is theoretically robust (infrared and collinear safe, identical procedure at parton level, at generator-particle level, and at reconstructed particle-level, etc). Experimentally, the jet clustering should be as fast as possible, should not be too sensitive to pile-up and underlying-event, should be able to separate nearby jets, and should not ignore large energy deposits. All of these effects will affect calorimeter-cell-only jet clustering differently compared with particle-flow jet clustering. Hence a variety of clustering algorithms need to be studied and evaluated in the context of particle-flow jets. Finally, the ability to "lock" (and "unlock") particles from being considered in the clustering process (such as identified leptons) needs to be developed.

Work to be done (and presented at the workshop):

  1. Develop "locking" functionality, which prevents certain particles from being considered in the clustering algorithm
  2. Compare different jet clustering algorithms at generator-level: cone algorithms (iterative, mid-point, seedless, etc), kT algorithm, Cambridge/Aachen, JADE, DURHAM, etc
  3. Compare the same above jet clustering algorithms for reconstructed particle-flow particles
  4. Study stability with respect to different calorimeter thresholds
  5. Study stability/performance due to different track efficiencies and fake rates
  6. Study separability of nearby jets (e.g. due to hard FSR or due to high prompt jet multiplicity)
  7. Study effect and stability of clustering in presence of nuclear interactions
  8. Study effect and stability of clustering in presence of underlying-event
  9. Study effect and stability of clustering in presence of pile-up (for different luminosities)
  10. Study effect of different fragmentation models

People involved (Volunteers welcome, there is work for many!) :

Jet Calibration

While most of the Jet energy scale calibration is expected to be determined from accurately calibrating the individual particles themselves, a small residual calibration will still likely be required. This residual calibration may arise from several possible effects, many of which can be identified and addressed now. For example, jets can be miscalibrated because the tracking becomes inefficient for very low ET jets and very high ET jets, the calorimeter thresholds affect low ET jets (mostly), the photon reconstruction and neutral hadron reconstruction become inefficient at low ET and at very high ET, and the presence of nuclear interactions affect the linking of tracks to calorimeter clusters. All of these effects will require new strategies for reconstructing and calibrating jets from particle flow. In addition, pile-up and underlying-event effects (which are "negative" corrections, while the others are "positive") will also have to be included, using the additional information available from the particle flow. Calibration constants must therefore depend not only on ET and eta but also on the charged particle, photon and neutral hadron content (and possibly muons, electrons, as well as jets identified with semi-leptonic decays), and the pT of the track experiencing nuclear interactions. The work can start simply at first, with Monte Carlo based corrections back to the final state particles of a jet as well as to the parton of a jet. However, it is also a long term effort which will ultimately require data drive strategies, many of which have an analog with the strategies being developed for calorimeter jets. All in all, the work to be done is not only to "develop corrections" but also think of a brand-new strategy suited for PF jets and the full information they carry.

Work to be done (not an exhaustive list and some of which should be presented at the workshop):

  1. Develop Monte Carlo based corrections back to the original parton of the jet
  2. Develop Monte Carlo based corrections back to the final state (visible) particles of a jet
  3. Develop strategies to correct for track inefficiencies (which can be large at low pT and at high pT)
  4. Develop strategies to correct for calorimeter thresholds (which primarily affect low ET jets)
  5. Develop strategies to correct for photon identification inefficiency or fakes (which can be large at low ET and at high ET)
  6. Develop strategies to correct for neutral hadron inefficiency or fakes (which can be large at low ET and at high ET)
  7. Develop strategies to correct for the presence of nuclear interactions
  8. Develop strategies to correct for pile-up: charged pile-up particles inside the acceptance of the tracker can be identified (as not belonging to the primary vertex) with some efficiency (the inefficiencies and fakes of which will need to be calibrated), however the effect due to neutral pile-up particles and pile-up particles beyond the acceptance of the tracker will also need to be studied.
  9. Develop strategies to correct effects due to the underlying-event (which depends on the effective pT of the event).
  10. Develop "effective" jet content corrections, which depend on the charged and neutral hadronic as well as the charged and neutral electromagnetic fractions of the jet.
  11. Develop "effective" relative jet corrections as a function of the jet pseudo-rapidity (such as dijet balancing), to remove any possible residual dependence on eta.
  12. Develop "effective" absolute jet corrections as a function of the jet momentum (or transverse momentum), to remove any possible residual dependence on pT.
  13. Develop corrections for different Jet Algorithms (sharing of energy between overlapping jets, etc)
  14. Determine if flavour dependent jet corrections are useful/necessary for particle flow jets.

People involved (Volunteers welcome, there is work for many!) : Michel Della Negra

MET determination

An accurate and precise determination of the missing energy of an event is one of the primary goals of the Particle-Flow Algorithm. By incorporating tracks and calorimeter clusters, redundant information is used by the Particle-Flow Algorithm to improve the response and resolution of the reconstructed energy for individual particles. However, because the missing energy of an event is a globally determined quantity which requires precise symmetric cancellations, any under-counting or over-counting of energy in the event will lead to fake missing energy. Hence a very accurate linking of calorimeter energy to track momentum is required. Additionally, any differences in energy resolution for particles in one hemisphere, compared with particles in the opposite hemisphere, will lead to fake missing energy. For all these reasons, it is important to study different physics processes (with/without intrinsic MET) and to systematically list as many effects which lead to fake missing energy, addressing them one-by-one (if they are not already—likely—addressed in another context). This work can easily and effectively be studied using the Fast Simulation (and later verified, where appropriate, using the Full Simulation). The PF-MET resolution and response should be determined as a function of the PF-MET and PF-SET of the event.

Work to be done (and presented at the workshop):

  1. Study back-to-back and single particle guns: pions, K0L's, electrons, photons
  2. Study events with no intrinsic MET as a function of PF-SET, such as QCD dijets (for different parton pT's) or Z' to dijets (for different Z' masses):
  3. Study events with intrinsic MET as a function of PF-MET and PF-SET, such as Z(nunu)+jet(s) (for different Z pT's)
  4. Study events with a high density of particles (ttbar, SUSY, etc)
  5. Scan above badly reconstructed events ( i.e. events which are in the tails)
  6. Study effect of Nuclear Interactions: Turn off/on Nuclear Interactions in the tracker and ECAL
  7. Study effect of improved satellite clustering
  8. Study effect of fake tracks: Use "only good" / "also bad" tracks
  9. Study effect of tracking efficiency: Use default / 3-step tracking
  10. Study effect of additional PF improvements
  11. Study effect of combining redundant information: determine PF-MET from PF-tracks alone and from PF-clusters alone
  12. Study PF-MET as an uncertainty weighted average of the energy from each of the input particles
  13. Study effect of different luminosities (PU conditions)
  14. Establish MET benchmark, for future reference and improvements

People involved: (Volunteers welcome! Needs more people.) Rick Cavanaugh

Tau reconstruction and tagging

Particle Flow reconstruction can bring notable improvement in the Tau reconstruction and identification:
  • Energy resolution: Already shown improvement up to energies of order 100 GeV
  • Eta-Phi resolution: having a more performant reconstruction of the jet direction will help in identifying tracks which are coming from the tau decay
  • Identification of particles inside the jet
    • easier suppression of electron contamination and their use to develop new tools for the tau identification
    • Better jet mass reconstruction

Tau identification tecniques can benefit from all the above points. As the number of informations available is becoming much higher than what used in the Calorimeter+Tracks identification, multivariate tecniques will become more and more important to take correctly into account all the relevant variables. At the moment only cut based analysis has been developed having mostly the same results as the calorimeter+tracks based identification. In order to develop new algorithms with MV analysis, we need:

  • C++ class which prepare the PFJet to be used by the Tagging algorithm, associating to the jets a set of selected particles: Charged Hadrons, Neutral Hadrons and Gamma candidates. The class is being written by Ludovic under the DataFormats/TauReco package: PFTauTagInfo.h
  • Produce a reco::PFTau object can be created with a set of variables which can be used to make the discrimination
    • Almost ready, need also to change a little the producer to start from the PFTauTagInfo as input
    • Different metrics implementations are available

* Write the code for the discrimination computation which may use the TMVA package. Actually this is completely missing * We need to add to the PFTau class a list TaggingVariables in order to use the TMVA package a la BTagging

* Create signal and background samples to test the performances.

People involved: (Volunteers welcome! Needs more people.) Ludovic is writing most of the code. Evan agreed to help after the 17th of Septmber. We need anyhow a third person who will take care of developping TMVA based tagging algorithm

Validation

It is important to study and track the improvements to the particle flow algorithms from software release to release, as well as to study and track the health of the particle flow from run to run in real data. For this, benchmark plots representing aggregate information (such as the tau, jet, and MET resolution and response plots) give a quick indication of the changes/health. In addition, more detailed information, such as the resolution (direction and energy) of individual particle species (charged/neutral hadrons, muons, electrons, unconverted/converted photons, V0s, pile-up particles, etc) will help to diagnose more particular improvements or possible problems. The goal is to develop a validation package which globally summarises statistical tests ("reference" versus "data") for high-level histograms (taus, jets, MET) and which allows one to "drill-down" to lower-level detailed histograms (individual PF-particles, PF-blocks, PF-clusters, etc) to understand the sources of possible differences with reference histograms.

Work to be done (and presented at the workshop):

  1. Incorporate the "Tau benchmark" into Validation/RecoParticleFlow package, and develop it to new energies and pseudo-rapidity ranges
  2. Develop and add Jet benchmark
  3. Develop and add MET benchmark
  4. Think of and include benchmarks for charged hadrons, neutral hadrons, muons, electrons, unconverted photons, converted photons, V0's , and pile-up particles
  5. Think of how to best represent statistical test information (comparing "reference" versus "data" histograms), providing the ability to navigate or "drill down" from histograms representing high-level information to histograms representing low-level details.
  6. Integrate with automatic software validation sub-system; publish results to web for easy access

People involved: Joanna Weng, Thomas Punz, Bobby Scurlock, Ron Remington, Michael Schmitt

Particle Flow in Fast Simulation

CSA07 Analyses

Review Status

Editor/Reviewer and date Comments
PatrickJanot - 12 Sept 2007 Preshower clustering
RickCavanaugh - 05 Sept 2007 Jet Calibration, Validation
PatrickJanot - 04 Sept 2007 V0 reconstruction, muons
RickCavanaugh - 04 Sept 2007 MET
PatrickJanot - 03 Sept 2007 Calibration and resolution maps, charged and neutral hadrons, merged photons, nuclear interactions, electrons
PatrickJanot - 03 Sept 2007 First draft of the page

Responsible: PatrickJanot
Last reviewed by:

Edit | Attach | Watch | Print version | History: r20 < r19 < r18 < r17 < r16 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r20 - 2008-03-07 - ColinBernet



 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    CMSPublic All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback