关键词: 有向超图, 链接预测, 超图卷积, 表示学习, 自适应卷积

Abstract: Abstract: Link prediction, as a typical problem in the field of graph learning, has always been a research hotspot. Although diverse solutions to this problem have been provided by graph neural networks, significant shortcomings in fully utilizing high-order and asymmetric information between vertices were identified due to the structural constraints of ordinary graphs. To address the above issues, a link prediction model based on directed hypergraph adaptive convolution was proposed. Firstly, the directed hypergraph structure was employed to more fully represent high-order and directional information between vertices, possessing advantages of both hypergraphs and directed graphs. Secondly, an adaptive information propagation method was adopted by directed hypergraph adaptive convolution to replace the traditional directional information propagation method in directed hypergraphs, thereby solving the problem of ineffective updating of embeddings for tail vertices of directed hyperedges and alleviating the problem of excessive smoothing of vertices caused by multi-layer convolution. Experimental results based on explicit vertex features on the Citeseer dataset show that the directed hypergraph adaptive convolution model achieves a 2.23 percentage point increase in the area under the ROC curve (AUC) and a 1.31 percentage point increase in average precision (AP) compared to the directed hypergraph neural network (DHNN) model in the link prediction task. Therefore, it can be concluded that this model adequately expresses the relationships between vertices and effectively improves the accuracy of the link prediction task.

Key words: directed hypergraph, link prediction, hypergraph convolution, representation learning, adaptive convolution