Rethinking information fusion: Achieving adaptive information throughput and interaction pattern in graph convolutional networks for collaborative filtering

Jiaxin Wu, Chenglong Pang, Guangxiong Chen, Jihong Wan, Xiaocao Ouyang, Jie Zhao

Abstract: Highlights•Heterogeneity in bipartite graphs is used for adaptive information fusion.•Adaptive quantification method is designed by information amount and node type.•Adaptive information throughput is achieved to improve GCN performance.•Different layer fusion methods are adapted to various user interaction patterns.•AdaptGCN achieves superior performance in both efficiency and performance.