GIPA: A General Information Propagation Algorithm for Graph Learning

Graph neural networks (GNNs) have been widely used in graph-structured data computation, showing promising performance various applications such as node classification, link prediction, and network recommendation. Existing works mainly focus on node-wise correlation when doing weighted aggregation of neighboring nodes based attention, dot product by the dense vectors two nodes. This may cause conflicting noise to be propagated information propagation. To solve this problem, we propose a General Information Propagation Algorithm (GIPA), which exploits more fine-grained fusion including bit-wise feature-wise correlations edge features their Specifically, calculates element-wise attention weights through multi-layer perceptron (MLP) representations edge; The is one-hot attribute for feature selection. We evaluate GIPA Open Benchmark proteins (OGBN-proteins) dataset Alipay Alibaba Group. Experimental results reveal that outperforms state-of-the-art models terms prediction accuracy, e.g., achieves an average ROC-AUC $$0.8917\,\pm \,0.0007$$ , better than all existing methods listed OGBN-proteins leaderboard.