RpHGNN

Source code and dataset of the paper "Source code and dataset of the paper Efficient Heterogeneous Graph Learning via Random Projection

Homepage and Paper

Homepage (RpHGNN): https://github.com/CrawlScript/RpHGNN
Paper: Efficient Heterogeneous Graph Learning via Random Projection

Requirements

Linux
Python 3.7
torch==1.12.1+cu113
torchmetrics==0.11.4
dgl==1.0.2+cu113
ogb==1.3.5
shortuuid==1.0.11
pandas==1.3.5
gensim==4.2.0
numpy==1.21.6
tqdm==4.64.1

Download Preparation

For HGB datasets (ACM, DBLP, Freebase, and IMDB)：

sh download_hgb_datasets.sh

For OAG-Venue and OAG-L1-Field, we follow NARS' data prepatation in https://github.com/facebookresearch/NARS/tree/main/oag_dataset. After generating *.pk and *.npy files, you have to:

put these files in the directory
rename graph_field.pk to graph_L1.pk

For OGBN-MAG, the code will automatically download it via the ogb package.

For OAG-Venue and OAG-L1-Field, we adhere to NARS' data preparation instructions found at https://github.com/facebookresearch/NARS/tree/main/oag_dataset. After generating *.pk and *.npy files, you should:

Place these files in the directory ./datasets/nars_academic_oag/.
Rename graph_field.pk to graph_L1.pk.

Run RpHGNN

You can run RpHGNN with the following command:

sh scripts/run_ACM.sh

sh scripts/run_DBLP.sh

sh scripts/run_Freebase.sh

sh scripts/run_IMDB.sh

sh scripts/run_OGBN-MAG.sh

sh scripts/run_OAG-Venue.sh

sh scripts/run_OAG-L1-Field.sh

Run RpHGNN for OGB Leaderboards (ogbn-mag)

To reproduce the results on the OGB Leaderboards (ogbn-mag), follow the steps below:

Preparing Pre-trained Embeddings (Optional):
- If the cache/mag.p file does not exist (embeddings pre-trained via LINE [1]), our code will automatically pre-train it and save the pre-trained embeddings in the specified path.
- Alternatively, if you'd prefer to skip the pre-training step, download the pre-trained embeddings mag.p directly from Google Drive and place it in the cache directory.
Execute the script:
```
sh scripts/run_leaderboard_OGBN-MAG.sh
```
This script will run the training and evaluation using random seeds from 0 to 9. The output for seed i will be saved in the file nohup_leaderboard_mag_i.out.

References:

[1] Jian Tang, Meng Qu, Mingzhe Wang, Ming Zhang, Jun Yan, and Qiaozhu Mei. "Line: Large-scale information network embedding." In Proceedings of the 24th international conference on world wide web, pp. 1067-1077. 2015.

Cite

If you use RpHGNN in a scientific publication, we would appreciate citations to the following paper:

@misc{hu2023efficient,
      title={Efficient Heterogeneous Graph Learning via Random Projection}, 
      author={Jun Hu and Bryan Hooi and Bingsheng He},
      year={2023},
      eprint={2310.14481},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}