You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 12 Next »

LCFIPlus

LCFIPlus is a reconstruction software which performs vertex finding, jet finding, and flavor tagging in one package.  It is implemented as a Marlin processor configurable with the XML steering file.

Reporting bugs & issues

If you have a SLAC confluence account, report it to the JIRA issue tracker:

https://jira.slac.stanford.edu/browse/LCFI

How to obtain LCFIPlus

The LCFIPlus package is included in the ILCSoft release.

The latest version can be downloaded from the DESY SVN repository:

svn co https://svnsrv.desy.de/public/marlinreco/LCFIPlus/trunk LCFIPlus

Discussions about the package happen on the mailing list lcfi-vertex AT desy DOT de

Algorithms

LCFIPlus contains the following algorithms:

  • PrimaryVertexFinder - searches for the primary vertex
  • BuildUpVertex - searches secondary vertices
  • JetClustering - jet clustering using vertex information
  • FlavorTag - prepares of flavor tagging input variables
  • MakeNtuple - produces ROOT files containing ntuples for training 
  • TrainMVA - performs TMVA training
  • ReadMVA - applies result of training
Example of PrimaryVertexFinder + BuildUpVertex
<processor name="MyLcfiplusProcessor" type="LcfiplusProcessor">

<!-- run primary and secondary vertex finders -->
<parameter name="Algorithms" type="stringVec"> PrimaryVertexFinder BuildUpVertex </parameter>

<!-- specify input collection names -->
<parameter name="PFOCollection" type="string" value="PandoraPFOs" />
<parameter name="PrimaryVertexCollectionName" type="string" value="PrimaryVertex" />
<parameter name="BuildUpVertexCollectionName" type="string" value="BuildUpVertex" />

<!-- parameters for primary vertex finder -->
<parameter name="PrimaryVertexFinder.TrackMaxD0" type="double" value="20." />
<parameter name="PrimaryVertexFinder.TrackMaxZ0" type="double" value="20." />
<parameter name="PrimaryVertexFinder.TrackMaxInnermostHitRadius" type="double" value="20." />
<parameter name="PrimaryVertexFinder.TrackMinVtxFtdHits" type="int" value="5" />
<parameter name="PrimaryVertexFinder.Chi2Threshold" type="double" value="25." />

<!-- parameters for secondary vertex finder -->
<parameter name="BuildUpVertex.TrackMaxD0" type="double" value="10." />
<parameter name="BuildUpVertex.TrackMaxZ0" type="double" value="20." />
<parameter name="BuildUpVertex.TrackMinPt" type="double" value="0.1" />
<parameter name="BuildUpVertex.TrackMaxD0Err" type="double" value="0.1" />
<parameter name="BuildUpVertex.TrackMaxZ0Err" type="double" value="0.1" />
<parameter name="BuildUpVertex.TrackMinTpcHits" type="int" value="20" />
<parameter name="BuildUpVertex.TrackMinFtdHits" type="int" value="3" />
<parameter name="BuildUpVertex.TrackMinVxdHits" type="int" value="3" />
<parameter name="BuildUpVertex.TrackMinVxdFtdHits" type="int" value="0" />
<parameter name="BuildUpVertex.PrimaryChi2Threshold" type="double" value="25." />
<parameter name="BuildUpVertex.SecondaryChi2Threshold" type="double" value="9." />
<parameter name="BuildUpVertex.MassThreshold" type="double" value="10." />
<parameter name="BuildUpVertex.MinDistFromIP" type="double" value="0.3" />
<parameter name="BuildUpVertex.MaxChi2ForDistOrder" type="double" value="1.0" />
<parameter name="BuildUpVertex.AssocIPTracks" type="int" value="1" />
<parameter name="BuildUpVertex.AssocIPTracksMinDist" type="double" value="0." />
<parameter name="BuildUpVertex.AssocIPTracksChi2RatioSecToPri" type="double" value="2.0" />

</processor> <processor name="MyLcfiplusProcessor" type="LcfiplusProcessor">
</processor>

Input variables for flavor tagging

Variables are computed for each jet.

name

description

trk1d0sig

d0 significance of track with highest d0 significance

trk2d0sig

d0 significance of track with second highest d0 significance

trk1z0sig

z0 significance of track with highest d0 significance (ordering by d0, not z0)

trk2z0sig

z0 significance of track with second highest d0 significance (ordering by d0, not z0)

trk1pt

transverse momentum of track with highest d0 significance

trk1pt_jete

trk1pt divided by the jet energy

trk2pt

transverse momentum of track with second highest d0 significance

trk2pt_jete

trk2pt divided by the jet energy

jprobr

joint probability in the r-phi plane

jprobz

joint probability in the z projection

vtxlen1

decay length of the first vertex in the jet (zero if no vertex is found)

vtxlen2

decay length of the second vertex in the jet (zero if number of vertex is less than two)

vtxlen12

distance between the first and second vertex (zero if number of vertex is less than two)

vtxsig1

decay length significance of the first vertex in the jet (zero if no vertex is found)

vtxsig1_jete

vtxsig1 divided by the jet energy

vtxsig2

decay length significance of the second vertex in the jet (zero if number of vertex is less than two)

vtxsig2_jete

vtxsig2 divided by the jet energy

vtxsig12

vtxlen12 divided by its error as computed from the sum of the covariance matrix of the first and second vertices, projected along the line connecting the two vertices

vtxdirdot1

the cosine of the angle between the momentum (computed as a sum of track four momenta) and the displacement of the first vertex

vtxmult1

number of tracks included in the first vertex (zero if no vertex is found)

vtxmult2

number of tracks included in the second vertex (zero if number of vertex is less than two)

vtxmult

number of tracks which are used to form secondary vertices (summed for all vertices)

vtxmass1

mass of the vertex computed from the sum of track four-momenta

vtxmasspc

mass of the vertex with minimum pt correction allowed by the error matrices of the primary and secondary vertices

vtxprob

vertex probability

vtxlen2

 

 

 

 

 

 

 

 

 

How to add a new variable

Each variable is defined inside FlavorTag.cc as a class inheriting from FTAlgo. The name of the variable is defined by providing a string argument to the constructor of the parent class. The variable class should implement the process() method which performs the variable computation. One can access the jet data through the member "_jet" and the event data through "_event".  The result should be stored in the member variable "_result". One can optionally implement the processEvent() method which is called once per every event. The class should be instantiated and added to the FTManager object inside the FlavorTag::init() method to make the variable known to LCFIPlus.

An example implementation of a variable is shown below.

// computes the number of vertices in a jet
class FtNvtx : public FTAlgo {class FtNvtx : public FTAlgo {
  public:
    FtNvtx() : FTAlgo("nvtx") {}
    void process() { _result = _jet->getVertices().size(); }
};

Using training results

Example of getting flavor tagging information inside a Marlin processor:

// get jet collection
LCCollection* colJet = evt->getCollection("VertexJets");

// get PIDHandler associated with the jet collection
PIDHandler pidh( colJet );

// get algorithm ID associated with LCFIPlus
int algo = pidh.getAlgorithmID( "lcfiplus" );

// get index number for flavor tagging
int ibtag = pidh.getParameterIndex(algo, "BTag");
int ictag = pidh.getParameterIndex(algo, "CTag");

// loop over jets to extract flavor tagging information
for(int i=0; i < colJet->getNumberOfElements(); i++) {
  ReconstructedParticle *part = 
    dynamic_cast<ReconstructedParticle*>( colJet->getElementAt( i ) );
  const ParticleID &pid = pidh.getParticleID(part, algo);
  cout << "btag = " << pid.getParameters()[ibtag] << endl;
  cout << "ctag = " << pid.getParameters()[ictag] << endl;
}
  • No labels