This repository is an implementation of GLEM - Learning on Large-scale Text-attributed Graphs via Variational Inference in ICLR 2023. Majority of the work is completed during Jianan's internship at Microsoft Research Asia.

The proposed GLEM framework trains GNN and LM separately in a variational EM framework: In E-step, an LM is trained towards predicting both the gold and GNN predicted pseudolabels; In M-step, a GNN is trained by predicting both the gold and GNN predicted pseudolabels.

The core packages are as below:

• python=3.8
• ogb=1.3.3
• numpy>=1.19.5
• dgl>=0.8.0
• pytorch=1.10.2
• pyg=2.0.3

To use our exact environment, one may install the environment by the following command:

conda env create -f environment.yml

Please look the details in Readme.md of each dataset inside the OGB folder.

Performance on ogbn-arxiv(10 runs):

Performance on ogbn-products(10 runs):

Performance on ogbn-papers100M(3 runs):

If you find our work useful, please consider citing our work:

@article{zhao2023glem,
  title={Learning on Large-scale Text-attributed Graphs via Variational Inference},
  author={Jianan Zhao and Meng Qu and Chaozhuo Li and Hao Yan and Qian Liu and Rui Li and Xing Xie and Jian Tang},
  journal={ICLR},
  year={2023}
}

The different models used for different datasets are referenced in the corresponding OGB subfolders