Online Bayesian Collaborative Topic Regression

TL;DR

This paper introduces an online Bayesian approach to collaborative topic regression that jointly optimizes matrix factorization and topic modeling in real-time, improving scalability and predictive performance for streaming data in recommender systems.

Contribution

It proposes a novel online Bayesian collaborative topic regression method that jointly optimizes PMF and LDA, addressing batch limitations and leveraging rating information more effectively.

Findings

Effective on real-world streaming data

Outperforms existing batch CTR algorithms

Scalable and efficient online learning

Abstract

Collaborative Topic Regression (CTR) combines ideas of probabilistic matrix factorization (PMF) and topic modeling (e.g., LDA) for recommender systems, which has gained increasing successes in many applications. Despite enjoying many advantages, the existing CTR algorithms have some critical limitations. First of all, they are often designed to work in a batch learning manner, making them unsuitable to deal with streaming data or big data in real-world recommender systems. Second, the document-specific topic proportions of LDA are fed to the downstream PMF, but not reverse, which is sub-optimal as the rating information is not exploited in discovering the low-dimensional representation of documents and thus can result in a sub-optimal representation for prediction. In this paper, we propose a novel scheme of Online Bayesian Collaborative Topic Regression (OBCTR) which is efficient and scalable for learning from data streams. Particularly, we {\it jointly} optimize the combined objective function of both PMF and LDA in an online learning fashion, in which both PMF and LDA tasks can be reinforced each other during the online learning process. Our encouraging experimental results on real-world data validate the effectiveness of the proposed method.