This code is a PyTorch implementation of our ICLR'24 paper "Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting". [arXiv]
🌟 Pathformer代码在阿里云仓库也进行同步更新：阿里云Pathformer代码链接
If you find this resource helpful, please consider to cite our research:
@inproceedings{chen2024pathformer,
author = {Peng Chen and Yingying Zhang and Yunyao Cheng and Yang Shu and Yihang Wang and Qingsong Wen and Bin Yang and Chenjuan Guo},
title = {Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting},
booktitle = {International Conference on Learning Representations (ICLR)},
year = {2024}
}
Pathformer, a Multi-Scale Transformer with Adaptive Pathways for time series forecasting. It integrates multi-scale temporal resolutions and temporal distances by introducing patch division with multiple patch sizes and dual attention on the divided patches, enabling the comprehensive modeling of multi-scale characteristics. Furthermore, adaptive pathways dynamically select and aggregate scale-specific characteristics based on the different temporal dynamics.
The important components of Pathformer: Multi-Scale Transformer Block and Multi-Scale Router.
To install all dependencies:
pip install -r requirements.txt
You can access the well pre-processed datasets from Google Drive, then place the downloaded contents under ./dataset
- Download datasets and place them under ./dataset
- Run each script in scripts/, for example
bash scripts/multivariate/ETTm2.sh
1, Transformers in Time Series: A Survey, in IJCAI 2023. [GitHub Repo]
@inproceedings{wen2023transformers,
title={Transformers in time series: A survey},
author={Wen, Qingsong and Zhou, Tian and Zhang, Chaoli and Chen, Weiqi and Ma, Ziqing and Yan, Junchi and Sun, Liang},
booktitle={International Joint Conference on Artificial Intelligence(IJCAI)},
year={2023}
}