# More recommender algorithms

I wanted to share some more insight into the algorithms we use at Spotify. One matrix factorization algorithm we have used for a while assumes that we have user vectors bf{a}_u and item vectors bf{b}_i . The next track i for a user is now given by the relation P(i \mid u) = exp(bf{a}_u^T bf{b}_i) / Z_u   Where Z_u is a normalization constant to make sure the probabilities add up to one. This essentially means we treat the tracks choices as outputs of a softmax. You can think of it as a series...