WWGAN (Worm Wasserstein GAN)

This is the official Pytorch implementation of WWGAN model presented in https://ieeexplore.ieee.org/document/9760052

Worm Wasserstein Generative Adversarial Network(WWGAN) is a Generative Adversarial Network (GAN) for time-series data augmentation. WWGAN builds upon two WGAN-GP by constructing the WGAN-GP into a recurrent structure like RNN to improve its data augmentation ability on time-series data.

WWGAN can generate synthetic time-series data that carry realistic intrinsic patterns with the original data and expands a small sample without prior knowledge or hypotheses. Which means it can augment one-dimensional time-series data.

The following is the architecture of WWGAN model.

Requirements

• Python 3.8.10
• Pytorch 1.9.0
• Numpy 1.21.2
• Matplotlib 3.5.1
• Pandas 1.2.4
• Scipy 1.6.2
• Seaborn 0.11.1

A virtual environment is recommended for running this project. The required dependencies are listed in requirements.txt.

Quick start

1. Run the WWGAN_toy.py to learn the toy dataset.
2. Run the WWGAN_verification.py to drow some figs to verify the augmentation results.

The real sample and fake(generated) data:

The fitting verification:

Customization

• The WWGAN model can be customized by modifying the model.py.
• The input time-series data can be changed into other datasets. The Hyperparameters should be modified accordingly.

Publication

If you found this code useful, please cite our paper:

• Title: Small Sample Reliability Assessment With Online Time-Series Data Based on a Worm Wasserstein Generative Adversarial Network Learning Method.

• Citation: Bo Sun, Zeyu Wu, Qiang Feng et al., "Small Sample Reliability Assessment With Online Time-Series Data Based on a Worm Wasserstein Generative Adversarial Network Learning Method," in IEEE Transactions on Industrial Informatics, vol. 19, no. 2, pp. 1207-1216, Feb. 2023, doi: 10.1109/TII.2022.3168667.

@ARTICLE{9760052,
  author={Sun, Bo and Wu, Zeyu and Feng, Qiang and Wang, Zili and Ren, Yi and Yang, Dezhen and Xia, Quan},
  journal={IEEE Transactions on Industrial Informatics}, 
  title={Small Sample Reliability Assessment With Online Time-Series Data Based on a Worm Wasserstein Generative Adversarial Network Learning Method}, 
  year={2023},
  volume={19},
  number={2},
  pages={1207-1216},
  doi={10.1109/TII.2022.3168667}}

Acknowledgements

This repository is based on the code published in WGAN-GP.