Towards Relation-centered Pooling and Convolution for Heterogeneous Graph Learning Networks

TL;DR

This paper introduces PC-HGN, a relation-centered graph neural network that improves heterogeneous graph learning by relation-specific sampling and cross-relation convolutions, outperforming existing models on real-world datasets.

Contribution

The paper proposes a novel relation-centered pooling and convolution approach for heterogeneous graphs, reducing customization and computational costs while enhancing structural encoding.

Findings

PC-HGN outperforms baseline models by up to 17.8% on real-world datasets.

Relation-specific sampling improves the encoding of structural heterogeneity.

Cross-relation convolutions enhance the model's ability to learn from diverse relation types.

Abstract

Heterogeneous graph neural network has unleashed great potential on graph representation learning and shown superior performance on downstream tasks such as node classification and clustering. Existing heterogeneous graph learning networks are primarily designed to either rely on pre-defined meta-paths or use attention mechanisms for type-specific attentive message propagation on different nodes/edges, incurring many customization efforts and computational costs. To this end, we design a relation-centered Pooling and Convolution for Heterogeneous Graph learning Network, namely PC-HGN, to enable relation-specific sampling and cross-relation convolutions, from which the structural heterogeneity of the graph can be better encoded into the embedding space through the adaptive training process. We evaluate the performance of the proposed model by comparing with state-of-the-art graph learning models on three different real-world datasets, and the results show that PC-HGN consistently outperforms all the baseline and improves the performance maximumly up by 17.8%.