The submission 6.09 is my best on public leaderboard without postprocessing, so I chose a submission with post processing and a submission without, my CV results were almost the same so I chose according to the public
My CV didn't change much either, in fact it was very difficult to get anything consistent, so I had to use as many submissions as I had.
I ended up making a similar choice, a more aggressive one with the best lb and one with a more robust model in terms of overfitting and which would have more chances in the private lb.
Thanks for the answer and congratulations on the 2nd place in the competition.
Maybe still arranging according to the your selected submissions?
Congratulations Belal_Emad
Was your choice of submission 6.09 because of robustness of cross validation?
Thank you!
The submission 6.09 is my best on public leaderboard without postprocessing, so I chose a submission with post processing and a submission without, my CV results were almost the same so I chose according to the public
My CV didn't change much either, in fact it was very difficult to get anything consistent, so I had to use as many submissions as I had.
I ended up making a similar choice, a more aggressive one with the best lb and one with a more robust model in terms of overfitting and which would have more chances in the private lb.
Thanks for the answer and congratulations on the 2nd place in the competition.
Which models did you use @rafael_zimmermann & @Belal_Emad & @data_style_bender? Good job & congratulations