The input variables for the latest track quality classifier (emulated version, runs with newKF emulation) changed. However, the variables set for the simulation version of the code did not get updated to reflect this change in the classifier. This PR updates the variables to the proper ones.
The changes in the variables are a consequence of the emulation. In emulation tanl is more readily available than eta and z0 is scaled by a non-base 2 value which makes it more costly to change back. Therefore, the bdt uses tanl and a scaled version of the z0 that are both easier to obtain.
PR validation:
I checked that the outputs from the MVA variable calculated by the cmssw code matched the output from the python version of the classifier that was used for the training of the bdt. The results here show perfect matching between the cmssw and python version of the classifier:
PR description:
The input variables for the latest track quality classifier (emulated version, runs with newKF emulation) changed. However, the variables set for the simulation version of the code did not get updated to reflect this change in the classifier. This PR updates the variables to the proper ones.
The changes in the variables are a consequence of the emulation. In emulation tanl is more readily available than eta and z0 is scaled by a non-base 2 value which makes it more costly to change back. Therefore, the bdt uses tanl and a scaled version of the z0 that are both easier to obtain.
PR validation:
I checked that the outputs from the MVA variable calculated by the cmssw code matched the output from the python version of the classifier that was used for the training of the bdt. The results here show perfect matching between the cmssw and python version of the classifier: