Recommender Systems and Graph Neural Networks

The core of recommender systems is to predict user’s preference over items, for example whether a user will buy a product or rate a movie etc. A major approach of recommender system design is collaborative filtering which uses past interactions between users and items to achieve the prediction. Graph neural networks are a type of deep learning method that achieves state-of-the-art performance in many deep learning tasks including link prediction.

Since the user-item interaction network has a graph structure, graph neural networks have been used as an ideal tool for collaborative filtering. Our research focuses on developing novel graph neural networks and their application in recommender systems.

Research areas

• Recommender systems
• Neural collaborative filtering
• Graph neural networks
• Graph spectral analysis
• Bipartite graph motifs

Current projects

• Motif-based graph neural networks
• Low-pass graph neural networks for collaborative filtering
• Deep learning for stock prediction

Research networks

• Professor Michael Sheng, Macquarie University, Australia
• Professor Jun Han, Swinburne University of Technology, Australia
• Professor Yanbo Han, North China University of Technology, China
• Professor Guiling Wang, North China University of Technology, China
• Professor Chengfei Liu, Swinburne University of Technology, Australia
• Professor Hong-Linh Truong, TU Wien, Austria
• Professor Lianghuai Yang, Zhejiang University of Technology, China