I think what they do is that they take 80% of the test data that we submit, and do the hamming loss scoring on that 80%. The final score is the hamming loss of the entire submission file we uploaded. So they won't need our code, as they have held back 20% of our submitted values which are included for the final leaderboard
I think they did run the models on the entire dataset(public and private leaderboard)
I’m not too sure though
IMO, We need to share the code first, so that they run on whole dataset.
Lets see what they(@zindi) have to say.
I think what they do is that they take 80% of the test data that we submit, and do the hamming loss scoring on that 80%. The final score is the hamming loss of the entire submission file we uploaded. So they won't need our code, as they have held back 20% of our submitted values which are included for the final leaderboard
I see, Thank You @allank.