Curriculum Learning for Debiased Recommendation with Explicit and Implicit Feedback

Abstract The recommender system (RS) has played an increasingly important role in Internet applications. Recent literature on RS mainly focused better fitting the user behavior data. However, data is observational, not experimental. This makes for a wide range of biases In this paper, we introduce novel framework to combine advantages both multi-task and curriculum learning debiased recommendation. Unlike existing methods that are limited specific feedback, our method follows unify explicit implicit feedback. And these two feedbacks learned manner by shifting from implicit. way, only use available information but also overcomes task-balancing problem learning. Extensive experiments have been conducted real-world datasets prove delivers state-of-the-art performance significantly improves debiasing ability recommendation model.