We have submitted all our predictions and got strange results on leadrboard 2. For leaderboard 1, we got very good results in both cross validation and leaderboard submissions (AUPR= 0.91 and AUC= 0.76). However, we get good results in cross validation but supper bad results on leaderboard 2, i.e. even 10% worse than a random prediction! It is more strange that before bug we got the second place for leareboard 2 and now the story is completely different. We are confused! any body has similar problem or can guess what is wrong with our predictions?
Thanks,
- Samad
Created by S J SAJA Suffered from same trouble, cross validation results are not consistent with current leaderboard results, subchallenge 1 though.
Drop files to upload
Is it possible to get a 40% gap between cross-validation and leaderboard scoring? Any idea from other groups? page is loading…