Path-aware Siamese Graph Neural Network for Link Prediction
This work addresses link prediction in graph networks, offering incremental improvements through a novel method for known bottlenecks in capturing node content and behavior.
The paper tackles link prediction by proposing a Path-aware Siamese Graph neural network (PSG) that captures node and edge features, including structure and relay paths, using a multi-task GNN with self-supervised contrastive learning. It achieves top 1 performance on ogbl-ddi and top 3 on ogbl-collab datasets.
In this paper, we propose a Path-aware Siamese Graph neural network(PSG) for link prediction tasks. First, PSG captures both nodes and edge features for given two nodes, namely the structure information of k-neighborhoods and relay paths information of the nodes. Furthermore, a novel multi-task GNN framework with self-supervised contrastive learning is proposed for differentiation of positive links and negative links while content and behavior of nodes can be captured simultaneously. We evaluate the proposed algorithm PSG on two link property prediction datasets, ogbl-ddi and ogbl-collab. PSG achieves top 1 performance on ogbl-ddi until submission and top 3 performance on ogbl-collab. The experimental results verify the superiority of our proposed PSG