TAEGAN

This repository contains the official implementation of the paper "TAEGAN: Revisit GANs for Tabular Data Generation", published at the 17th Asian Conference on Machine Learning (ACML 2025), PMLR volume 304. [Paper]

Pre-requisites

1. Python>=3.10.
2. Python package torch, tqdm, sklearn, numpy.

Execution

Prepare Data

python run.py -o OUT_DIR prepare -s DATA_FILE

What the prepare step does is described in run.py under prepare function. Relevant code can be easily reconstructed from open-source CTAB-GAN+ implementation. We decided not to expose our processing code because we are using code source from enterprise, which is not supposed to be exposed, The part of code actually called from this private code essentially does the same thing as CTAB-GAN+ data preprocessing, with only code structure and efficiency improvements.

Train Model

python run.py -o OUT_DIR train -b BATCH_SIZE -e EPOCHS -w WARMUP_EPOCHS

The trained model weights are found in OUT_DIR/generator.pt and OUT_DIR/discriminator.pt.

Sample Tensor

python run.py -o OUT_DIR sample -n N_ROWS -b BATCH_SIZE

The sampled tensor is found in OUT_DIR/sampled.pt. Note that the output of this step is still a transformed tensor with one-hot and VGM data encoding.

Recover Sampled Data

python run.py -o OUT_DIR recover

This is the inverse step of prepare, with its description also found inside run.py, in function recover. Since this is the inverse step of prepare, which is dependent on the data processing module, this part of code is also not exposed, but would be very easy to reconstruct. The output should then be saved as a csv file, which is the actually generated tabular data.

Example

A running example is provided in diabetes-demo using the diabetes dataset from OpenML. Content inside is the outcome of prepare step.

One can directly run the following to get the sampled tensor:

python run.py -o diabetes-demo train
python run.py -o diabetes-demo sample -n 256

Citation

If you use this code in your research, please cite our paper:

@InProceedings{pmlr-v304-li26c,
  title = {TAEGAN: Revisit GANs for Tabular Data Generation},
  author = {Li, Jiayu and Zhao, Zilong and Yee, Kevin and Javaid, Uzair and Sikdar, Biplab},
  booktitle = {Proceedings of the 17th Asian Conference on Machine Learning},
  pages = {670--685},
  year = {2025},
  editor = {Lee, Hung-yi and Liu, Tongliang},
  volume = {304},
  series = {Proceedings of Machine Learning Research},
  month = {09--12 Dec},
  publisher = {PMLR},
  pdf = {https://raw.githubusercontent.com/mlresearch/v304/main/assets/li26c/li26c.pdf},
  url = {https://proceedings.mlr.press/v304/li26c.html}
}

License

This code is released for non-commercial research and academic purposes only. See LICENSE for details.