PrivacyGANs

• ArXiv: Can we Generalize and Distribute Private Representation Learning?
• Github: PrivacyGANs
• Data: Drive Link

Instructions

• Datasets have to be downloaded individually as per regulations and copyrights.
• Drive contains the model checkpoints, training histories, and corresponding plots.
• Data is in the same directory structure as required by project (paste in corresponding folders).
• docker integration is used to reduce the overhead of setting up environment.
• users are welcome to use non-docker environments on their own.
• prepopulated hyperparameters and training logs as well as pretrained models are made available for evaluation.

Docker Setup

To build the docker image

• replace gpu with cpu in docker-dl-setup/docker-compose.yml in case the system has no gpu
• script to build the docker image

cd docker-dl-setup
docker-compose build

To run the docker container

./run-docker.sh

To enter the docker container

docker exec -it eigan_devel bash

Training

• all scripts are run from *.sh files in scripts folder
• change the hyperparameters, as in example scripts
• run the scripts inside the docker container

sh scipts/<mimic/mnist/titanic>/<script-name>.sh

Source folder executions

cd src
sh sh/<script-name>.sh <expt-name>

Comparison

• comparison scripts need editing of python scripts
• replace the names of the pre-populated training histories with the newly generated training histories after training to generate new plots and analysis.

Citation

If you find the repository or the paper useful, please cite the following paper

@InProceedings{azam2022can,
  title={{ Can we Generalize and Distribute Private Representation Learning? }},
  author={Azam, Sheikh Shams and Kim, Taejin and Hosseinalipour, Seyyedali and Joe-Wong, Carlee and Bagchi, Saurabh and Brinton, Christopher},
  booktitle={Proceedings of The 25th International Conference on Artificial Intelligence and Statistics},
  pages={11320--11340},
  year={2022},
  editor={Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel},
  volume={151},
  series={Proceedings of Machine Learning Research},
  month={28--30 Mar},
  publisher={PMLR},
  pdf={https://proceedings.mlr.press/v151/shams-azam22a/shams-azam22a.pdf},
  url={https://proceedings.mlr.press/v151/shams-azam22a.html}
}