Self-restrained contrastive enhanced network for graph structure learning

Nan Jia, Xiaolin Tian, Ting Yang, Shuo Li, Licheng Jiao

Abstract: Highlights•We propose a GNN model for learning and optimizing graph structure.•Multiple graph structures fuse with their multi-order information.•The graph structure is adjusted and optimized by the self-restrained module.•The contrast enhancement module is adopted to get the distinguishable feature.•The proposed GNN model alleviates the over-smoothing problem of GNNs.