Next Steps of PrKalmanFilter & TrackMasterFitter Development
Following up on discussion in WP2 and TrackFit meetings.
We want to summarise and continue discussion of how to proceed with the development of the new PrKalmanFilter.
These major points come to mind:
1. More Validation of PrKalmanFilter
-
resolution + pull plots for all output states -
Momentum resolution vs p -
Move Cheated MM providers to PrKalmanFilter -
Integrate into HLT2 selections -
Check selection efficiencies and produce mass plots, e.g. decays with Ks. -
Fix issue seen where RICH1 Cherenkov resolution shows a significant degradation with the new fitter.
2. Alignment
Scenario A: use TrackMasterFitter for the alignment.
Scenario B: use the PrKalmanFilter, this needs a few changes. Thanks to @wouter we already know a few things that need to be implemented (this is considered the preferred scenario (in March '21))
-
Implement the possibility to calculate alignmentDerivatives -
Implement the Classical Smoother -
Setup a workflow that tests this somehow to even start developing and testing this - additional details can be found in this helpful note written by @wouter
3. Monitoring
-
Most monitoring code which considers (unbiased) residuals uses the FitNodes --> likely best to make a monitored version of the fit that adds these to the output.
4. "Optional" functionality
-
Include the progress in supporting multi-dimensional measurements (by @jdevrie)
cc @wouter @graven @ldufour @sstahl
Please feel free to edit the above to add items to the lists.
Edit:
With Test_PrKalman.py it possible to run the new and old fitter in a job and run TrackResChecker.
Just do: Moore/run gaudirun.py Test_PrKalman.py
If you want to just do quick and dirty plots of the histos produced by the above job you can do use TrackResChecker_plot.py Just do: (assuming `histo2.root is in the same dir)
source /cvmfs/sft.cern.ch/lcg/views/LCG_97a/x86_64-centos7-gcc9-opt/setup.sh
python TrackResChecker_plot.py