I was wondering the reasoning behind having to submit the probability scores for the validation set as we have the labels for these scores.
Cheers
Posted by: HarryCoppock @ March 16, 2021, 5:28 p.m.While Test AUC is used to rank order the teams, displaying the VAL AUCs gives an additional dimension to compare the teams. All team use the same VAL folds to report the scores on hence, we get to know how well the teams are generalizing their approaches across different data subsets. With a small dataset, the intention is to get multiple points of comparison across teams. Hope this clarifies.
Posted by: dicova2021 @ March 17, 2021, 6:40 a.m.