Recommender Systems
Two fitted offsets that deliver two thirds of the accuracy gain before any latent factor is learned, a model 1.28 times worse for the users who have told it least, and the blind spot that opens when a system only ever sees ratings for what it chose to show.
Sign in to take quizzes, earn XP, and unlock stages as you reach 90% mastery.
The Baseline That Is Hard to Beat
25 min · 100 XPA rating matrix that is 82% empty, the two offsets that explain most of what is in it, and the latent factors that earn the remaining third of the gain.
Open lesson →Sign in to take the 3-question quiz.
Cold Start and Who the Model Fails
28 min · 110 XPThe error is not spread evenly: it is 1.28 times worse for the users with fewest ratings, which is exactly the group a recommender most needs to convince.
Open lesson →Sign in to take the 3-question quiz.
The Loop That Teaches Itself
30 min · 120 XPOnly 30% of the catalogue ever appears in anyone’s top ten, with no explicit popularity term in the model at all, and after six rounds of learning from its own recommendations the system is 1.14 times worse exactly where it never looked.
Open lesson →Sign in to take the 3-question quiz.