/MGAN

Source code for the paper MGAN: Training Generative Adversarial Nets With Multiple Generators

Primary LanguagePythonMIT LicenseMIT

Mixture Generative Adversarial Nets (MGAN)

This TensorFlow code implements an example of MGAN for the CIFAR-10 dataset, presented in the paper "MGAN: Training Generative Adversarial Nets with Multiple Generators" accepted at the 6th International Conference on Learning Representations (ICLR 2018).

The code is tested on Linux-based operating system with Python 3.6, TensorFlow 1.2.1. The data for this example can be downloaded here.

Run the model using this command:

python main.py

Please kindly look at the file main.py for hyperparameter arguments.

Citation

Quan Hoang, Tu Dinh Nguyen, Trung Le, Dinh Phung. "MGAN: Training Generative Adversarial Nets with Multiple Generators". International Conference on Learning Representations, 2018.

Bibtex

@inproceedings{
        hoang2018mgan,
        title={{MGAN}: Training Generative Adversarial Nets with Multiple Generators},
        author={Quan Hoang and Tu Dinh Nguyen and Trung Le and Dinh Phung},
        booktitle={International Conference on Learning Representations},
        year={2018},
        url={https://openreview.net/forum?id=rkmu5b0a-},
        }