Ranking: Feature Engineering and Learning-to-Rank
~40 min read
Scoring the candidate set with rich features and deep models: cross-feature engineering, CTR regression, and learning-to-rank losses.
Ranking: Feature Engineering and Learning-to-Rank is a Pro topic
Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.
Key points
- •Cross features (user × item attribute combinations) are the most powerful ranking signals — they capture interaction effects no user or item feature alone can express
- •Pointwise CTR regression is the most common ranking loss in production; pairwise and listwise losses are theoretically better but harder to scale
- •Wide & Deep combines memorization (sparse cross-features → linear) with generalization (dense embeddings → MLP) — both are needed for reliable production ranking