ASIAM-HGNN: Automatic Selection and Interpretable Aggregation of Meta-Path Instances for Heterogeneous Graph Neural Network

keywords: Graph neural networks, meta-paths, network representation learning, heterogeneous graph
In heterogeneous information network (HIN)-based applications, the existing methods usually use Heterogeneous Graph Neural Networks (HGNN) to handle some complex tasks. However, these methods still have some shortcomings: 1) they manually pre-select some meta-paths and thus some important ones are missing, while the missing ones still contains the information and features of the node in the entire graph structure; and 2) they have no high interpretability since they do not consider the logical sequences in an HIN. In order to deal with them, we propose ASIAM-HGNN: a heterogeneous graph neural network combined with the automatic selection and interpretable aggregation of meta-path instances. Our model can automatically filter important meta paths for each node, while preserving the logical sequence between nodes, so as to solve the problems existing in other models. A group of experiments are conducted on real-world datasets, and the results demonstrate that the models learned by our method have a better performance in most of task scenarios.
reference: Vol. 42, 2023, No. 2, pp. 257–279