Note on an item-item recommender system using matrix factorisation

Collaborative filtering for product recommendations tends to focus on user-item recommendations – given a user, try to find the products they might like based on the products they have already interacted with.  While personalisation like this is often preferable, sometimes we would just like to know what products to recommend if a user lands on a given product page.  There is far less written about this in the literature, so I would like to share some thoughts (and maths).

Continue reading