The code of paper ”Robust Heterogeneous Graph Neural Networks against Adversarial Attacks“ (AAAI2022).

Here we show the attack process in "attack_HAN.ipynb and attack_HAN_RoHe.ipynb".

To clearly compare the robustness of HAN and our HAN-RoHe, we also provide their visualizations in Figure attack_HAN.png and attack_HAN_RoHe.png. *For simplicity, in the code we take 100 target nodes as example, which can be change to 500 in codes.

Run attack_HAN.ipynb and attack_HAN_RoHe.ipynb

-torch 1.3.1

-dgl

-deeprobust

-sklearn

More details will be published soon (^▽^)