Cluster error tuning

• Automatized search (choose half widths around optimum and submit next iteration)
• Adjusted search metric: Correctly attached non-fake * average efficiency (primaries)
  • Efficiency improved slightly
• Adjusted network input: All cluster flags, invCharge, invAvgCharge, mult
  • That didn't change much...
• Added new scaling parameters for chi cuts in tracking
  
  
  • First search: scaleChiY1, scaleChiZ1, nnScaleClusterErrorY, nnScaleClusterErrorZ

• Next step: Need new (more realistic) simulation
• Tighten cuts on training data further to have a better cluster error estimate