Memorization Rejection GAN (MRGAN)

This repo implements the memorization rejection technique to prevent training sample memorization in GANs proposed in our paper Reducing Training Sample Memorization in GANs by Training with Memorization Rejection. The codebase is largely derived from the StudioGAN repo, which I have no ownership over. Please refer to the StudioGAN repo for how to run experiments. Arguments for controlling memorization rejection can be found in src/main.py.

Please cite our work if you find this repo useeful

@misc{https://doi.org/10.48550/arxiv.2210.12231,
      doi = {10.48550/ARXIV.2210.12231},
      url = {https://arxiv.org/abs/2210.12231},
      author = {Bai, Andrew and Hsieh, Cho-Jui and Kan, Wendy and Lin, Hsuan-Tien},
      keywords = {Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences},
      title = {Reducing Training Sample Memorization in GANs by Training with Memorization Rejection},
      publisher = {arXiv},
      year = {2022},
      copyright = {Creative Commons Attribution 4.0 International}
}