spliew / pearl

Data Synthesis via Private Embeddings and Adversarial Reconstruction Learning

PEARL

• This repository contains source codes to reproduce the results in PEARL: Data Synthesis with Private Embeddings and Adversarial Reconstruction Learning, published as a conference paper at ICLR 2022: https://openreview.net/forum?id=M6M8BEmd6dq

Structure

• image: contains the code for image dataset experiment. To run, use main.py. To evaluate, use eval.py.
• tabular: contains the code for tabular dataset experiment. To run or evaluate, use run.py.

Dependencies

Version numbers are based on our machine and may need not to be matched exactly.

scikit-learn 0.23.1
pytorch 1.5.0
torchvision 0.6.0
matplotlib 3.2.1
seaborn 0.10.1
sdgym 0.2.2

License

This implementation is licensed under the Apache License 2.0.

Acknowledgement

Our implementation refers to the source code from the following repositories:

• Technical Report: Relational Data Synthesis using Generative Adversarial Networks: A Design Space Exploration
• DP-MERF
• GAN Metrics Pytorch; licensed under the Apache License 2.0.