AIC-GNN: Adversarial information completion for graph neural networks.

•We propose a novel information completion strategy for GNNs to learn low-degree node representation better.•We realize the information completion strategy as the general extension of GNNs in an end-to-end manner, named AIC-GNN, with rich experimental results.•We propose novel dual node embedding alignment mechanisms as supplements for AIC-GNN, which could effectively facilitate simulating the node missing information distribution.