Performance Evaluation

Under:

Note:  This is a rather old study that I found on BEMC public AFS space and ported into Drupal.  I don't know who conducted it or when.  -- A. Kocoloski

On the subject of reconstruction made by cluster finder and matching, evaluation has been developed  to determine how good is the reconstruction, efficiency, purity, energy and position resolution of reconstructed particles originally generated by simulation data.

Cluster finder is being evaluated at the moment using single particle events, which favors the evaluation of efficiency and purity. In this case, we can define efficiency and purity for single particle events as:

  • efficiency - ratio between the number of events with more than 1 cluster and the total number of events.
  • purity - ratio between the number of events with only 1 cluster and the number of events with at least one cluster.

There are other quantities that could be used for evaluation. They are:

  • energy ratio - ratio between the cluster energy and the geant energy.
  • position resolution - difference between cluster position and particle position.

All these quantities can be studied as a function of the cluster finder parameters,  mSizeMax, mEnergySeed and mEnergyThresholdAll. The results are summarized below.

BTOW Clusters

mEnergyThresholdAll evaluation

Nothing was done to evaluate this parameter.

mEnergySeed evaluation

The purity as a function of mEnergySeed gets better and the efficiency goes down as this parameter increases.  Some figures (the others parameters were kept as minimum as possible) are shown for for different values of mSizeMax for single photons in the middle of a tower with pt = 1,4,7 and 10 GeV/c.

The following plots were generated using energy seeds of 0.1 GeV (left), 0.5 GeV (middle), and 1.5 GeV (right).  Full-size plots are available by clicking on each image:

Eta dfference

Phi difference


Number of clusters



Energy ratio

mSizeMax evaluation

Nothing was done to evaluate this parameter.

BSMD Clusters

mEnergyThresholdAll evaluation

Nothing was done to evaluate this parameter.

mEnergySeed evaluation

The purity as a function of mEnergySeed gets better and the efficiency goes down as this parameter increases.  Some figures (the others parameters were kept as minimum as possible) are shown for for different values of mSizeMax for single photons in the middle of a tower with pt = 1,4,7 and 10 GeV/c.

The following plots were generated using energy seeds of 0.05 GeV (left), 0.2 GeV (middle), and 0.8 GeV (right).  Full-size plots are available by clicking on each image:

Eta Difference (BSMDE only)


Eta difference RMS as a function of photon energy for different seed energies:


Phi Difference (BSMDP only)

Phi difference RMS as a function of photon energy for different seed energies:

Number of clusters for BMSDE

Number of clusters for BSMDP:

BSMDE efficiency and purity as a function of seed energy

BSMDP efficiency and purity as a function of photon energy for different seed energies

mMaxSize evaluation

The efficiency and purity as a function of mSizeMax get better as this parameter increases but for values greater than 4 there is no difference. Some figures (the others parameters were kept as minimum as possible) are shown for for different values of mSizeMax for single photons in the middle of a tower with pt = 5 GeV/c.