Converting to top-n recommendation

[joohyelim]

I am very impressed about the usefulness of AutoInt to my personal project.
I'm wondering if there is any handy way of converting this AutoInt to top-n recommendaiton?
Would it be possible?

[Songweiping]

I think you can first pair the target user with all possible items and then score each <TargetUser, item> pair with AutoInt model. To get top-n recommendations, just rank all candidate items based on predicted scores.