rogerni / mole

MoLE (AISTATS 2024)

This is the official implementation of the paper "Mixture-of-Linear-Experts for Long-term Time Series Forecasting". [arXiv] [PMLR]

Requirements

Please refer to the requirements.txt file for the required packages.

Datasets

All datasets we used in our experiments (except Weather2K) are available at this Google Drive's shared folder. These datasets were first provided in Autoformer. Please download the datasets and put them in the dataset folder. Each dataset is an .csv file.

Weather2K dataset is available at this GitHub repository.

Usage

Main Experiments

To run the main experiments, please run the following command:

scripts/run_all_3_seeds.sh

This scripts sequentially runs the main experiments on all datasets with 3 different random seeds. The results will be saved in the logs folder.

Additional Experiments

The repository has been cleaned up for easier usage. If you want to run ablation experiments, please refer to earlier commits.

Acknowledgement

We thank the authors of the following repositories for their open-source code or dataset, which we used in our experiments:

https://github.com/zhouhaoyi/Informer2020

https://github.com/cure-lab/LTSF-Linear

https://github.com/plumprc/RTSF

https://github.com/bycnfz/weather2k

Citation

If you find our work useful, please consider citing our paper using the following BibTeX:

@inproceedings{ni2024mixture,
  title={Mixture-of-Linear-Experts for Long-term Time Series Forecasting},
  author={Ni, Ronghao and Lin, Zinan and Wang, Shuaiqi and Fanti, Giulia},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  pages={4672--4680},
  year={2024},
  organization={PMLR}
}