The repo is the official implementation for the paper: "CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting" (IJCAI 2025)
CASA regards CNN Autoencoder-based Score Attenton improving channel-wise tokenization and shows Model Agnostic Feature including Computational Efficiency.
The main result of CASA is as the following:
- Install Pytorch and necessary dependencies.
pip install -r requirements.txt
- Train and evaluate the model. We provide all the above tasks under the folder ./scripts/. You can reproduce the results as the following examples:
# ECL dataset : Multivariate forecasting with CASA
bash ./scripts/long_term_forecast/ECL_script/CASA.sh
If you find this repo helpful, please cite our paper.
@misc{lee2025casacnnautoencoderbasedscore,
title={CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting},
author={Minhyuk Lee and HyeKyung Yoon and MyungJoo Kang},
year={2025},
eprint={2505.02011},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2505.02011},
}