• graph
• link prediction
• node classification
• predictive approach
• self-supervised learning

Self-supervised learning on graph structured data has seen rapid growth especially centered on augmentation-based contrastive methods. In this thesis, we will propose a self-supervised predictive approach that aims to reconstruct the information of a node using its neighbors. We demonstrate the effectiveness of our approach in both edge-level and node-level task. Our model shows competitive performance in node classification. In link prediction task, our model outperforms self-supervised model from literature.