(I might not make it today: Presentation in the TPC weekly meeting)

Minor updates today

• Testing multiclass training for classification and regression, limited successs so far
• After some talking with Mesut about neutron capture and common mode
  • Checked if NN can be applied on cluster level (i.e. fill input with native CoG positions and charges instead of digits): Works and gives also reasonably good results -> Can be applied offline in order to reduce clusters in the current data
• Temporary fix of MC labels for NN clusters by building clusters with GPU cluster finder and storing MC labels -> needs some investigation but recovers fake-rate in QA plots
• Otherwise improved QA plots for CHEP