I just found the docker container doesn’t have Catboost installed. But I don’t have too much time to test the docker container. So is it possible to provide Catboost installation as quickly as possible using other ways? Because I have only one submission chance left.
@glipstein @cszc just give a pull request, please have a check, thanks!
@LastRocky just merged. Shouldn’t take more than 20 minutes for the container to be built and published.
Thanks! So will that be OK if I submit now(submission need Catboost)? Just want to make sure it won’t fail, since it’s my last submission. @cszc
@LastRocky Yep, it should work! And if it doesn’t, a failed submission won’t count toward your limit so it should still be fine.
Hi @LastRocky, the competition is over now so I think we can talk. I’ve seen you are scoring good and I wonder if you’ve been able to get a decent score with CatBoost? I gave it a try but, even optimising and trying with some feature engineering, the score was far from good (much better with keras).
Hi @adalseno, in my offline experiments, CatBoost is performing much better than LightGBM, but not as good as my NN model I designed, which is a LSTM based model. But unluckily I have some stupid bugs in my last submission, which was supposed a much better result than my second last submission (but didn’t…).