Hi Would it be possible to provide teams with the AUROC and AUPR when their classifier is run against the training set? This would give teams a better sense of whether they are under/overfitting without exposing any sensitive data. Thank you

Created by Jeff Green jeff_green
Thanks for your comments and for considering the idea! Yes, all teams would likely benefit, and hopefully it would help with the overall success of the project.
Hi @jeff_green, Thank you for suggesting this. My experience thus far has been that the temptation to tune models to the "leader board" is very strong without some feedback of this nature.
Hi @jeff_green, I think that's a good recommendation. I've started discussions internally to see whether this is feasible and advisable. Thanks for bringing this to our attention. @trberg

AUROC, AUPR from training set page is loading…