Recommender systems play a vital role in modern web services. In a typical recommender system, we are given a set of observed user-item interaction records and seek to uncover the hidden behavioral patterns of users from these historical interactions. By exploiting these hidden patterns, we aim to discover users’ personalized tastes and recommend them new items. Among various types of recommendation methods, the latent factor collaborative filtering models are the dominated ones. In this paper, we develop a unified view for the existing latent factor models from a probabilistic perspective. This is accomplished by interpreting the latent representations being drawn from their corresponding underlying representation distributions, and feeding the representations into a transformation function to obtain the parameters of the observation sampling distribution. The unified framework enables us to discern the underlying connections of different latent factor models and deepen our understandings of their advantages and limitations. In particular, we observe that the loss functions adopted by the existing models are oblivious to the geometry induced by the item-similarity and thus might lead to undesired models. To address this, we propose a novel model— textsf{SinkhornCF}—based on Sinkhorn divergence. To address the challenge of the expensive computational cost of Sinkhorn divergence, we also propose new techniques to enable the resulting model to be able to scale to large datasets. Its effectiveness is verified on two real-world recommendation datasets.

The Web Conference is announcing latest news and developments biweekly or on a monthly basis. We respect The General Data Protection Regulation 2016/679.