Closed mpearmain closed 9 years ago
does it help? i implemented dropout in difacto (factorization machine). it only helps on some datasets. i'm wonder if it's necessary for the much simpler linear method
On the Avazu Kaggle competition i benefited from dropouts with multiple passes of the data.
thanks for the explanation. can i ask you to do me a favor to test the dropout if it's implemented?
No problem at all, I have been doing some more testing today on private data to verify the results are scalable
On Tue, 28 Jul 2015 at 17:48 Mu Li notifications@github.com wrote:
thanks for the explanation. can i ask you to do me a favor to test the dropout if it's implemented?
— Reply to this email directly or view it on GitHub https://github.com/dmlc/wormhole/issues/16#issuecomment-125676382.
Have you thought about adding dropouts as an enhancement? Similar to the below https://github.com/bingzhengwei/ftrl_proximal_lr