<ul><li>Graph Neural Networks (GNN) face catastrophic forgetting, hindering knowledge preservation.</li><li>Non-Exemplar methods like Prototype Replay (PR) address memory issues in GNN.</li><li>Prototype Contrastive Learning (PCL) shows reduced drift compared to conventional PR.</li><li>Instance-Prototype Affinity Learning (IPAL) is proposed for Non-Exemplar Continual Graph Learning (NECGL), outperforming existing methods.</li></ul>

Instance-Prototype Affinity Learning for Non-Exemplar Continual Graph Learning

Discover more