Member-only story
Bayesian cold-start recommender
I’ve spent much of the past year prototyping, productionising, and developing the efficient learning algorithm for BBC Bitesize. Our ambitions for the Bayesian framework that underpins the Bitesize quiz have always gone beyond just a quiz.
Once we got the initial prototype of the quiz up and running we turned our attention to diversification, and specifically towards adapting the algorithm to a cold-start recommender for BBC News and Sport.
Cold Start Recommender
The term “cold start” refers to the lack of data associated with new users to a product. How do you recommend something to someone when you know nothing about them?
Probably the most famous recommenders are based on the matrix factorisation method that won the Netflix Prize. While these models are highly performant, they require vast quantities of data about a user’s previous viewing in order to become effective.
The efficient learning algorithm is, first and foremost, a cold-start algorithm. BBC Bitesize wanted the quiz to work with non-signed in users that make up most of their audience. Our solution was a Bayesian algorithm that updates its…
