Recommender Systems
Recommender systems are among the highest-leverage ML in production. Learn content-based and collaborative filtering, similarity and matrix factorization, the implicit-vs-explicit feedback distinction, ranking metrics (precision@k, NDCG), the cold-start problem, and modern hybrid and neural approaches.
- Chapter 01
Foundations
3 lessons - 01 Why recommenders matter Netflix, Spotify, Amazon, TikTok — how do they decide what to put in front of you, and why is that decision worth billions?
- 02 The utility matrix How recommender systems represent 'who likes what' — and why filling in the blanks is the whole problem.
- 03 Content-based filtering Recommend new items nobody has rated yet — by representing every item as a feature vector and matching it to what the user has already loved.
- Chapter 02
Collaborative Filtering
4 lessons - 04 User-based collaborative filtering How 'people similar to you liked X' works — finding a neighborhood of like-minded users and using their ratings to predict yours.
- 05 Item-based collaborative filtering User-based CF breaks at scale. Item-based CF flips the axis — compare items instead of users — and lets you precompute similarities offline. This is the engine behind Amazon's 'customers who bought this also bought.'
- 06 Similarity metrics Cosine, Pearson, and Jaccard: which similarity measure to use in collaborative filtering, and why choosing the wrong one gives you wrong neighbors.
- 07 Matrix factorization (SVD, ALS) Learn how the Netflix Prize winners modeled taste as latent factors by approximating the sparse utility matrix as a product of two thin matrices — the idea behind modern recommendation engines.
- Chapter 03
Production Concerns
4 lessons - 08 Implicit vs explicit feedback Almost no one leaves star ratings, but everyone clicks and watches. Learn how to build recommenders from behavioral signals alone using preference-confidence modeling.
- 09 Evaluating recommenders (precision@k, NDCG) Why RMSE is the wrong target for recommender systems, and how precision@k, recall@k, MAP, and NDCG measure what users actually experience.
- 10 The cold-start problem A new user just signed up and a new item just launched — neither has any interaction history. What on earth do I recommend?
- 11 Hybrid & neural recommenders How production systems at YouTube/Netflix scale go beyond a single collaborative-filtering model — combining content-based and CF signals as hybrids, and replacing hand-crafted similarity with learned embeddings, Neural CF, and two-tower retrieval.
- End of section 0 / 11 complete
Make it stick — pass every quiz.
Each lesson has a short quiz at the bottom. Passing the quiz is what marks the lesson complete and counts toward your certificate.
Section complete 11 / 11 lessonsNice work — you finished Recommender Systems.
Certificates are earned per learning path, not per section. Here's where this section takes you:
Pick a learning path to start working toward a certificate.
Recommender Systems — frequently asked questions
Straight answers to the questions people ask most about recommender systems.
What is the difference between content-based and collaborative filtering?
Content-based filtering recommends items similar to what you already liked using item features such as genre, tags, or text, while collaborative filtering uses the behavior of many users — 'people like you also liked' — without needing item features. Content-based handles brand-new items well, collaborative filtering captures taste patterns features can't, and hybrid systems combine both.
Read the lessonWhat is the cold-start problem in recommender systems?
Cold start is when you can't make good recommendations because there is no interaction history yet — for a new user, a new item, or a brand-new system. The fixes are content-based filtering and metadata for new items, onboarding preferences and popularity fallbacks for new users, and deliberate exploration to gather data.
Read the lessonWhat is matrix factorization in recommender systems?
Matrix factorization approximates the sparse user-item rating matrix as the product of two smaller matrices of latent factors, one for users and one for items, so a predicted rating is the dot product of a user vector and an item vector. These learned factors capture hidden taste dimensions and powered the Netflix Prize-winning approaches.
Read the lessonWhy is RMSE a poor metric for recommender systems?
Because recommendation is a ranking problem — users only ever see the top few items — so a rating-prediction error like RMSE doesn't measure whether the right items reached the top. Use top-k ranking metrics such as precision@k, recall@k, and NDCG, and always compare against a simple popularity baseline.
Read the lesson