Heterogeneous Graph Neural Networks (HGNNs) are a class of powerful deep learning methods widely used to learn representations of heterogeneous graphs. Despite the fast development of HGNNs, they still face some challenges such as over-smoothing, and non-robustness. Previous studies have shown that these problems can be reduced by using gradient re...

Unifying gradient regularization for Heterogeneous Graph Neural Networks