sp8rks opened this issue 3 years ago · 1 comments
was train/val/test just random or was clustering etc applied to ensure more generalizable models?
It was done randomly. Indeed, we acknowledge there can be potential data leakage.