I am motivated to design graph learning algorithms that are robust to in-the-Wild distribution shifts in real-world applications. Specifically, I work on:

Principles: transferable pre-training of GNNs (EGI), handling localized training data (SR-GNNs), domain adaptation via optimal Transport (GDOT)
Applications: entity alignment on unknown types(CG-Align), name disambiguation in academia network (GAND)and etc.