Scalable Recommendation with Hierarchical Poisson Factorization

We develop hierarchical Poisson matrix factor-ization (HPF), a novel method for providing users with high quality recommendations based on implicit feedback, such as views, clicks, or purchases. In contrast to existing recommen-dation models, HPF has a number of desirable properties. First, we show that HPF more accu-rately captures the long-tailed user activity found in most consumption data by explicitly consider-ing the fact that users have finite attention bud-gets. This leads to better estimates of users ’ la-tent preferences, and therefore superior recom-mendations, compared to competing methods. Second, HPF learns these latent factors by only explicitly considering positive examples, elimi-nating the often costly step of generating arti-ficial negative examples when fitting to implicit data. Third, HPF is more than just one method— it is the simplest in a class of probabilistic models with these properties, and can easily be extended to include more complex structure and assump-tions. We develop a variational algorithm for ap-proximate posterior inference for HPF that scales up to large data sets, and we demonstrate its per-formance on a wide variety of real-world recom-mendation problems—users rating movies, lis-tening to songs, reading scientific papers, and reading news articles.