Personalized Ranking on Poisson Factorization.
Journal
Proceedings of the 2018 SIAM International Conference on Data Mining, SDM 2018, May 3-5, 2018, San Diego Marriott Mission Valley, San Diego, CA, USA.
Pages
720-728
Date Issued
2018
Author(s)
Abstract
Matrix factorization (MF) has earned great success on recommender systems. However, the commonly-used regression-based MF is not only sensitive to outliers but also unable to guarantee that the predicted values are in line with the user preference orders, which is the basis of common measures in recommender systems, e.g., nDCG. To overcome this drawback, we propose personalized ranking on Poisson factorization (PRPF), which utilizes the posteriori based on pair-wise learning to rank instead of the classical regression-based ones. Since the posteriori that combines learning to rank and Poisson factorization does not follow the conjugate prior relationship, we estimate variational parameters approximately and propose two optimization approaches based on variational interference. Due to the combination, PRPF not only preserves user preference but also performs well on a sparse matrix. In the experiment, we show that PRPF outperforms the state-of-the-art methods and achieves promising results for recommendation tasks.
Type
conference paper