Improving Graph Neural Network Models in Link Prediction Task via A Policy-Based Training Method

Graph neural network (GNN), as a widely used deep learning model in processing graph-structured data, has attracted numerous studies to apply it the link prediction task. In these studies, observed edges are utilized positive samples, and unobserved randomly sampled negative ones. However, there problems sampling samples. First, some missing that existing network. Second, can be easily distinguished from edges, which cannot contribute sufficiently Therefore, using directly samples is difficult make GNN models achieve satisfactory performance To address this issue, we propose policy-based training method (PbTRM) improve quality of proposed PbTRM, sample selector generates state vectors determines whether select them We perform experiments with three on two standard datasets. The results show PbTRM enhance