Overlapping Community Detection with Graph Neural Networks
May 28, 2020 ยท View on GitHub
Pytorch implementation of the Deep Variational Generative Overlapping Community Detection
Usage
The main algorithm and other utilities(metrics, evaluation .etc) are implemented in the embedding_cd package that can be installed as
python setup.py install
The main analysis and visualization code are in Jupyter notebooks
run
pip install -r requirements.txt
first.
Requirements
numpy>=1.16.4
pytorch>=1.2.0
scipy>=1.3.1
dgl>=0.42
...
Some modifications in DGL version:
- append
F.reluafter logits BatchNormafter GCNLayeynocd.utils.l2_reg_loss(model, scale=args.weight_decay)rather than setting weight_decay inAdamdropout -> Relu -> BN