Difference: Log-JunghwanGoh-200801 (13 vs. 14)

20080130

• Memory error
  • memory allocation exception (vector::reserve)

• DQM still not working
• First result (Z to MuMu? sample & CMSSW_1_7_5)
  • Too low efficiency!! why?
  • simEta, Phi, pT cut?

• e-mailed to developers

20080128-29

• Conversion to TFileService? is done !
• Submitted to crab

20080126

• DQM still not working
• Alternative way to store histograms
  • Added option to choose TFileService? for histoManager
  • TFileService? works!!!
  • Needed creation of Multi-depth directory - not supported by TFileService?::mkdir()

20080125

• Testing & Debugging new RecoMuonValidator?
  • DQM not working (histogram booking is good, but cannot fill entry.. strange)
  • std::alloc error sometimes.
  • Adding TFile and TFileService? method to store histograms

20080123

• Implementation is almost done!
  • unmatched simTrack handling - big one to draw efficiency: done yesterday midnight
  • pair<T::const_iterator> to pair<T&> : 현재의 C++표준에 의해 reference형의 reference를 취할 수 없게 되어 있음.
• Compiles well, test jobs
  • castor is down? : On @ 3 pm

20080122

• loop 에서 iterator를 도는 중에 container를 수정하면 iterator가 무효화되어 crash된다!

20080117-18

• updated/cvs commit RecoMuonValidator? to support Multi-track event
  • Algorithm is : Simple DeltaR? matching with cuts
• Implementing table based matching algorithm
  • Basic structure is done
  • code snipping from TrackAssociation? (AssociationByChi2?, etc)

20080115-16

Sim - Reco matching table

• use map<double, pair<TP,Mu>> instead of map<pair<TP,MU>, double>
  • this is not symantic, but no problem to use
  • automatically sorted at insertion time : fast and super-easy
• multimap is better because quality value can be same (Chi2, nHit, etc)
  • note that if same quality but no common TP/Mu then this is not a duplicated pairs.
• Adam already removed FitSlicesYTool?

20080114

Sim - Reco matching issue

• Non-physical result of RecoMuonValidator?
  • efficiency is almost flat and >1 in some cases
  • this is caused by duplication of reco-tracks in sim to reco matching
  • Can be avoided requiring minimal cuts, but is not an ultimate solution.
• Matching quality table : simTrack ID : recoTrack ID => ΔR/Χ²/Hits, etc
  • How to choose best matching pairs?
    • Sorting sim-reco pairs by qualities
      • Qualifier should be normalized
      • Can be nice if considers duplication
    • Traditional sim-reco, reco-sim matching with consideration of duplication
      • Order of key can be an issue : which one to choose if Q[sim1:rec1] = Q[sim2:rec1] ?
  • Is it a "Confusion matrix" in CS : No --
    • Not a square matrix in general : efficiency ≠ 1, fakerate ≠ 0
    • Extend by inserting "fake" and "lost" species?
• How to make N-argumentd function
  • overload comma operator to return the same class : is it possible?

20080110-13

• Multi-track extension of RecoMuonValidatior? gives non-physical result : duplication of reco-tracks
  • Solution : use trackAssociator / or new method?

20080109

• Extending RecoMuonValidator? to Multi-track event
  • Small modification is enough to implement, except matching part
  • Testing.

20080108

• Fini les vacance - Back to work !

-- JunghwanGoh - 08 Jan 2008