Cluster error parametrization

• Tried with cluster error dump within the track interpolation + further cleanup
• 2D grid searches for best scaling parameters for scaleY*err2Y, scaleZ*err2Z
• Stuck to learning MSE, ((track_pos - cluster_pos)^2 - track_cov), where track_cov is now taken in intermediate interpolation stage

Combining several grid runs

Cubic spline interpolation

• This is not the full picture -> Very sensitive to long / short tracks

• A "good" configurations (scaleY, scaleZ): (0.01, 0.1)

• 6% track reduction (default reconstruction: 388k tracks on the given dataset)

Getting close but not there yet