As we are only submitting CSV files with model predictions over testing datasets, this is more like a validation dataset right? During the last days of the competition, the model will be tested against unseen data, right?
Thanks
@air_hack If I’m understanding correctly, I think you’re mostly right on both accounts.
we are only submitting CSV files with model predictions over testing datasets
Yes, this is true. Leaderboard scores (private and public) are based off the test labels.
During the last days of the competition, the model will be tested against unseen data, right?
Yes, top models will be tested against an unseen data set. We are calling this a validation set, though I understand how that can be confusing. This validation will occur after the competition has ended but before winners are announced.
1 Like
Hi @cszc,
do unseen data mean unseen dates and times or unseen grid ids?
Thanks in advance!!!
Andrea
Hi @andreaRapuzzi - Just unseen dates and times.
2 Likes