Dear participants,
In case you have noticed that the AUC is very low for some data sets, this is normal, as we are using a normalized AUC, see (Sect. IV.C):
http://www.causality.inf.ethz.ch/AutoML/automl_ijcnn15.pdf
This is particularly true for RL, which is a sort of data set that will be present for sure in the final phase,
Best
HJ
Posted by: hugo.jair @ Jan. 17, 2018, 7:12 p.m.