GANE: A Generative Adversarial Network Embedding

This is the codes and data for GANE model in our paper "GANE: A Generative Adversarial Network Embedding".

If you would like to acknowledge our efforts, please cite the following paper:

@article{hong2019gane,
title={GANE: A Generative Adversarial Network Embedding},
author={Hong, Huiting and Li, Xin and Wang, Mingzhong},
journal={IEEE transactions on neural networks and learning systems},
year={2019},
publisher={IEEE}
}

Prerequisites

python 2.7

tensorflow

cPickle

numpy

multiprocessing

Usage

There are four files:

• utils.py: Utile functions for data preparing;
• dis_model.py: The discriminator of GANE;
• gen_model.py: The generator of GANE;
• gane.py: The trainer to minimax our discriminator and generator.

How to run

python gane.py --emb_dim 128 --epochs=150 --suffix 128d --init_lr_gen 1e-5 --init_lr_dis 1e-5

Note

• The output files (learned embeddings) will be stored in the output-suffix directory during training process, if --suffix suffix.