Distributed Recommendation Systems: Survey and Research Directions

With the explosive growth of online information, recommendation systems have become essential tools for alleviating information overload. In recent years, researchers have increasingly focused on centralized recommendation systems, capitalizing on the powerful computing capabilities of cloud servers and the rich historical data they store. However, the rapid development of edge computing and mobile devices in recent years has provided new alternatives for building recommendation systems. These alternatives offer advantages such as privacy protection and low-latency recommendations. To leverage the advantages of different computing nodes, including cloud servers, edge servers, and terminal devices, researchers have proposed recommendation systems that involve the collaboration of these nodes, known as distributed recommendation systems. This survey provides a systematic review of distributed recommendation systems. Specifically, we design a taxonomy for these systems from four perspectives and comprehensively summarize each study by category. In particular, we conduct a detailed analysis of the collaboration mechanisms of distributed recommendation systems. Finally, we discuss potential future research directions in this field.