This repository is the implementation of causal GNN. To solve the problem of generalization in GNN. The SCM model we define is as follows:
z is the intention variable, x and xn represent the characteristics of the central node and its neighbor nodes respectively, sn is structure information. The implementation process is shown in the following figure:- python 3.9
- pytorch 1.10
To train the model(s), run this command:
python train.py