The repo is the official implementation for the paper: Population Aware Diffusion for Time Series Generation.
Population-aware Diffusion for Time Series (PaD-TS) is a new TS generation model that better preserves the population-level properties. The key novelties of PaD-TS include 1) a new training method explicitly incorporating TS population-level property preservation, and 2) a new dual-channel encoder model architecture that better captures the TS data structure.
Environment setup.
$ conda env create --name PaD-TS --file=PaD-TS.yml
$ conda activate PaD-TS
Running experiment
$ python run.py -d {name} >& results/{name}.txtTS generation results with generation length 24 for Sines, Stocks, and Energy datasets. Bold font (lower score) indicates the best performance.
Long TS Generation Results on Energy dataset. Bold font (lower score) indicates the best performance.
If you find this repo useful, please cite our paper!
@article{li2025population,
title={Population Aware Diffusion for Time Series Generation},
author={Li, Yang and Meng, Han and Bi, Zhenyu and Urnes, Ingolv T and Chen, Haipeng},
journal={arXiv preprint arXiv:2501.00910},
year={2025}
}
Thanks for the open sources papers listed below which PaD-TS is build on.
https://github.com/openai/improved-diffusion/tree/main
https://github.com/thuml/iTransformer
https://github.com/facebookresearch/DiT