Hello all,
Congrats to the organisers for putting out and administering the contest and congrats to all the participants for their submissions!
With respect to the final leaderboard entries, I think it would be very interesting and enlightening if you also published ROC and PR AUC achieved by the submissions. This would give the community an even better understanding of what was achieved.
Kind regards
K. Vougas
Hi everyone,
First, thanks to organizers for all their work, I really enjoyed the competition!
Second, as an extension of @deepmed’s proposal: I know the competition data is going to be openly published soon, could you also publish not only additional metrics but the test set itself & corresponding labels and the probabilities outputs from LB solutions? That would be really helpful for error analysis
Thanks,
BR, Mikhail
Hi @deepmed and @mkotyushev, we will be open-sourcing the code for the winners as well as writing a blog post that will have some more interpretable metrics to give a better sense of how top solutions performed. This should help provide more context to the scores on the leaderboard. Thanks for the great engagement with the competition!
Hi @emily,
Thanks for your response. Will you also provide access to the test set?
@deepmed The open dataset to be released by VisioMel will include both train and test set data.