Diff-MTS: Temporal-Augmented Conditional Diffusion Model For Industrial Time Series

Source codes for the paper "Diff-MTS: Temporal-Augmented Conditional Diffusion-Based AIGC for Industrial Time Series Toward the Large Model Era": Diff-MTS [IEEE Transactions on Cybernetics] by Lei Ren, Haiteng Wang, Yuanjun Laili.

Diff-MTS is a novel diffusion-based AIGC model tailored for industrial multivariate time series (MTS). It leverages temporal augmentation and adaptive diffusion techniques to generate high-quality synthetic data, addressing challenges in industrial data generation, including data scarcity, unstable training in GANs, and complex temporal dependencies.

Usage

1. Clone this repository:

   git clone https://github.com/your-username/diff-mts.git
   cd diff-mts

2. Install

   pip install -r requirements.txt

3. Training the Model Train the Diff-MTS model using the following command:

   python MainCondition.py --epoch 50 --dataset FD001 --lr 2e-3 --state all --model_name DiffUnet --T 500 --window_size 48 --sample_type ddpm --input_size 14

Citation

If you find this code helpful, please cite our paper:

@article{ren2024diff,
  title={Diff-MTS: Temporal-Augmented Conditional Diffusion-Based AIGC for Industrial Time Series Toward the Large Model Era},
  author={Ren, Lei and Wang, Haiteng and Laili, Yuanjun},
  journal={IEEE Transactions on Cybernetics},
  year={2024},
  publisher={IEEE}
}

Ren L, Wang H, Laili Y. Diff-MTS: Temporal-Augmented Conditional Diffusion-Based AIGC for Industrial Time Series Toward the Large Model Era[J]. IEEE Transactions on Cybernetics, 2024.

Acknowledgment

Thanks for the lucidrains/denoising-diffusion-pytorch project for their contributions to this project.