csetraynor/multi-categorical-gans

Code for the paper "Generating Multi-Categorical Samples with Generative Adversarial Networks"

IMPORTANT UPDATE!

Please consider checking the code for my new work Improving Missing Value Imputation with Deep Generative Models.

Multi-Categorical GANs

Code for the paper Generating Multi-Categorical Samples with Generative Adversarial Networks

Pre-requisites

The project was developed using python 3.6.7 with the following packages:

• future==0.17.1
• numpy==1.16.0
• scikit-learn==0.20.2
• scipy==1.2.0
• torch==1.0.0

Installation with pip:

pip install -r requirements.txt

Contents

• Datasets
  • Synthetic data generation
  • US Census 1990
• Methods
  • ARAE and MC-ARAE
  • MedGAN and MC-MedGAN
  • MC-Gumbel
  • MC-WGAN-GP
• Metrics

Changelog

• 2019-01-28: changed to Python 3 as suggested (and still compatible with 2.7 ... I hope).
• 2018-07-25: now we use WGAN-GP for ARAE following the author updates.