Nice writeup, jhoon! In case anyone else was searching his/her eyes trying to figure out where the "likes" field is actually sanitized out from the the trainin

Nice writeup, jhoon! In case anyone else was searching his/her eyes trying to figure out where the "likes" field is actually sanitized out from the the training data set, it's the item.pop("likes") inside the function "is_liked": https://github.com/jeremydhoon/hashkernel/blob/master/fblearn.py#L53 https://github.com/jeremydhoon/hashkernel/blob/master/fblear... ... Maybe I like long names too much, but my vote would be something longer like "check_if_liked_and_remove_likes" since "is_liked" sounds like a predicate that won't be modifying its input.