Semantic-Specific Hierarchical Alignment Network for Heterogeneous Graph Adaptation

Node classification has been substantially improved with the advent of Heterogeneous Graph Neural Networks (HGNNs). However, collecting numerous labeled data is expensive and time-consuming in many applications. Domain Adaptation (DA) tackles this problem by transferring knowledge from a label-rich domain to a label-scarce one. However the heterogeneity and rich semantic information bring great challenges for adapting HGNN for DA. In this paper, we propose a novel semanticspecific hierarchical alignment network for heterogeneous graph adaptation, called HGA. HGA designs a sharing-parameters HGNN aggregating path-based neighbors and hierarchical domain alignment strategies with the MMD and L-1 normalization term. Extensive experiments on four datasets demonstrate that the proposed model can achieve remarkable results on node classification.