Looks like I did not add the kin fit confidence level to the features files. The feats files DO include kinfit chi2 and ndf, which I will use to calculate chi2/ndf (adding to pipeline now...), but this might not give the same separation as cl... Will check back.
UPDATE: chisq/ndf seems to do the job. If either it or cl are NOT included, the fits do not converge in 200 epochs -- only overfitting happens. INTERESTING.
No comments:
Post a Comment