Vishnu V. Zutshi for the NICADD/NIU group Potpourri Vishnu V. Zutshi for the NICADD/NIU group
Cluster Separability AHCal Meeting,DESY,Oct03 10/28/2003
Separability For any two classes or clusters the classification rule can basically be stated as: P(c1|x) > P(c2|x) The separability of two clusters or classification error probability depends on the difference between P(c1|x) and P(c2|x) Many measures can be constructed to quantify this AHCal Meeting,DESY,Oct03 10/28/2003
For Example….. For best separability clusters should have small within class variance and large between class distance. J = trace{Sw-1 Sm} where Sw = S PiSi Si is the covariance matrix for cluster ci Sm is the covariance matrix w.r.t. the global mean AHCal Meeting,DESY,Oct03 10/28/2003
Track Parameters AHCal Meeting,DESY,Oct03 10/28/2003
‘Separability’ vs. Distance AHCal Meeting,DESY,Oct03 10/28/2003
Another Measure…. B = a*(mi-mk)T *({Si+Sk}/2)-1 *(mi-mk) + b*ln{(|(Si+Sk)/2|)/sqrt(| Si || Sk |)} m = mean and S = covariance the first term gives the separation due to mean difference the second term due to covariance difference AHCal Meeting,DESY,Oct03 10/28/2003
‘Separability’ vs. Distance AHCal Meeting,DESY,Oct03 10/28/2003
Ratio of measures (m x m/n x n) AHCal Meeting,DESY,Oct03 10/28/2003
Ratio of measures (m x m/n x n) 3x3/1x1 5x5/3x3 AHCal Meeting,DESY,Oct03 10/28/2003
‘Resolution’ 3x3 1x1 5x5 AHCal Meeting,DESY,Oct03 10/28/2003
Test-beam Geometry Implemented in G4 HCal Tail-catcher ECal AHCal Meeting,DESY,Oct03 10/28/2003
G4 Setup GEANT 4.5.2.p01 Mokka physics lists For convenience General Particle Source linked in Aggregate hits by cell with track contributions ASCII output format (others can be put in) AHCal Meeting,DESY,Oct03 10/28/2003
Data Format EVENT evt_no ECAL no_hits layer_no … x y z cell_edep abs_edep ntrak track_no track_edep … HCAL … TAILCATCHER AHCal Meeting,DESY,Oct03 10/28/2003
HCal Face & Side View AHCal Meeting,DESY,Oct03 10/28/2003
Charged Pion Event AHCal Meeting,DESY,Oct03 10/28/2003
Tail-catcher AHCal Meeting,DESY,Oct03 10/28/2003
Extruder Line AHCal Meeting,DESY,Oct03 10/28/2003
Extruded Strip w/ 10 holes AHCal Meeting,DESY,Oct03 10/28/2003
10-hole Die AHCal Meeting,DESY,Oct03 10/28/2003
Tail-catcher Next Steps Steel from Fermilab ? Need to have a stable design to go hunting With the simulation now ready this should happen soon AHCal Meeting,DESY,Oct03 10/28/2003