• This repository is the implementation of CommunityGAN:

CommunityGAN: Community Detection with Generative Adversarial Nets

• data/: graph and community data
• pre_train/: pre-trained vertex embeddings

  Note: the dimension of pre-trained vertex embeddings should equal n_emb in src/CommunityGAN/config.py

• results/: evaluation results and the learned embeddings of the generator and the discriminator
• src/: source codes for CommunityGAN and pre-train model

The input file for CommunityGAN (data/community_detection/*_train.txt): An undirected graph in which vertex IDs start from 0 to N-1 (N is the number of nodes in the graph). Each line contains two node IDs indicating an edge in the graph.

The input file for the pre-train model (data/community_detection/*_agm.txt): Similar to the input file for CommunityGAN. The only difference is that in this file one edge need occur twice: node1 node2 and node2 node1

The community file (data/community_detection/*.sampled.cmty.txt): Each line means a community and the numbers indicate the vertices in the community.

The code of CommunityGAN has been tested running under Python 3.6.1, with the following packages installed (along with their dependencies):

• tensorflow == 1.6.0
• numpy == 1.12.1
• scipy == 1.1.0

The basic usage of CommunityGAN is as follow:

cd src/CommunityGAN
python community_gan.py

The parameters for CommunityGAN can be changed by editing src/CommunityGAN/config.py or passing through the command line. An example of running CommunityGAN on the three datasets are written in scripts/run.py, which can be called by:

cd scripts
python run.py

Get into the src/PreTrain directory and use make command to compile. Has been tested on Ubuntu with g++ 5.3.0, and on Windows with MinGW-w64 5.3.0.

A basic usage example of the pre-train model has been written in scripts/prepare_pretrain_embedding.py. The following commands can be used to re-prepare the pre-train embeddings for CommunityGAN on the three datasets:

cd scripts
python prepare_pretrain_embedding.py