Two recommenders in succession: The first one filters the item pool down to thousands in a highly efficient way; the second one ranks the remaining items and provides the resulting top-10 list to the user.

Two recommenders in succession: The first one learns rules from context and collaborative data; the second one uses these together with a content-based filtering component to make recommendations.

A weighted hybrid of two recommenders with either (β₁ = 0 and β₂=1) or (β₁=1 and β₂=0), depending on some external criterion.

Single-branch Embedding Networks for Recommendation

Content-boosted Collaborative Filtering

Several recommendation lists are shown to the user, next to each other (slate recommendation).