I. Introduction
Recommendation systems can help users discover items that they are interested through learning the user preference from the history records, item attributes and interaction context. In the era of big data, recommendation systems are becoming an important component of online platforms. From the perspective of rating type, recommendation algorithms can be divided into explicit recommendation and implicit recommendation [1]. Implicit feedbacks include not only ratings but also clicking, collection and play count etc. Implicit data is easier to get than explicit ratings, so the implicit feedback recommendations are attracted more attentions in recent years.