This repository contains the source code of our NeurIPS 2022 paper: "Self-supervised Heterogeneous Graph Pre-training based on Structural Clustering".
See the paper at OpenReview: https://openreview.net/forum?id=fBU4qsM6Fkf.
The other preprocessed datasets are too large to be posted on GitHub. They can be downloaded from Baidu cloud drive: https://pan.baidu.com/s/1_EemCZ4XOiIX07JemSTpaA?pwd=shgp.
For the Att-HGNN module of SHGP, we adopt our previously proposed ie-HGCN as its base encoder. Please refer to our ie-HGCN repository: https://github.com/kepsail/ie-HGCN.
python == 3.8.8
scipy == 1.6.2
torch == 1.9.1
numpy == 1.20.1
scikit-learn == 0.24.1
GPU: GeForce GTX 1080 Ti
CPU: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
Go to ./code, and execute the following command:
python main.py