Feature recommendation strategy for graph convolutional network

Graph Convolutional Network (GCN) is a new method for extracting, learning, and inferencing graph data that builds an embedded representation of the target node by aggregating information from neighbouring nodes. GCN decisive classification link prediction tasks in recent research. Although existing performs well, we argue current design ignores potential features node. In addition, presence with low correlation to nodes can likewise limit learning ability model. Due above two problems, propose Feature Recommendation Strategy (FRS) this paper. The core FRS employ principled approach capture both node-to-node node-to-feature relationships encoding, then recommending maximum possible replacing low-correlation features, finally using features. We perform clustering task on three citation network datasets experimentally demonstrate improve challenging relative state-of-the-art (SOTA) baselines.