Overview of DSGC: Deep Signed Graph Clustering

Installation

We have tested our code on Python 3.6.13 with PyTorch 1.8.0, PyG 1.8.1 and CUDA 12.1. Please follow the following steps to create a virtual environment and install the required packages.

Clone the repository:

git clone xxx
cd DSGC

Create a virtual environment:

conda create --name dsgc python=3.6.13 -y
conda activate dsgc

Install dependencies:

pip install torch==1.8.0+cu111 torchvision==0.9.0+cu111 torchaudio==0.8.0 -f https://download.pytorch.org/whl/torch_stable.html
pip install torch-scatter==2.0.6 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu111.html
pip install torch-sparse==0.6.9 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu111.html
pip install torch-cluster==1.5.9 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu111.html
pip install torch-spline-conv==1.2.1 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu111.html
pip install torch-geometric==1.6.3
pip install -r requirements.txt

Reproduce Results

We provide the source code to reproduce the results in our paper. The results of DSGC can be reproduced by running main.py.

To train labeled datasets:

python main.py --N [node_number] --p [edge_probability] --K [cluster_number] --eta [flip_probability] --delta_p [pos. threshold] --delta_n [neg. threshold] --m_p [add pos. edges] --m_n [add neg. edges]

To train unlabeled datasets:

python main.py --dataset [dataset_name]

The pos. threshold is the positive threshold for select noisy positive edges, and neg. threshold is the negative threshold for select noisy negative edges. Positive edges are added between two reachable nodes along m_p positive walks. Negative edges are added between two reachable nodes along m_p negative walks. Simple test on the signed stochastic block model SSBM (N=1000, p=0.01, K=5, eta=0.01). Dataset_name can be choosen from sp1500 and rainfall.

Current Authors

If you have problems with the code please contact:

xxx: ...@...
xxx: ...@...