I'm performing repeated 10-fold cross-validation in order to validate the predictive accuracy of an ML model.
the following paper https://www.hpl.hp.com/techreports/2009/HPL-2009-359.pdf is suggesting to average the AUCs of each fold within the 10-fold cross-validation.
Now my question is: I derive then 10 different AUCs (one for each repetition). Should I average the 10 different AUC in order to derive a summary AUC for the model?
I searched extensively papers or editorial material but it seems there is not so much literature about AUC computation within repeated cross-validation