This is a implementation of ST-Norm. The implementations of backbone Wavenet is cited from the published resource.
Python 3.7
Numpy >= 1.17.4
Pandas >= 1.0.3
Pytorch >= 1.4.0
python main.py --mode train --snorm 1 --tnorm 1
model: backbone architecture (wavenet / tcn / transformer).
snorm: whether use spatial normalization.
tnorm: whether use temporal normalization.
dataset: dataset name.
version: version number.
hidden_channels: number of hidden channels.
n_pred: number of output steps.
n_his: number of input steps.
n_layers: number of hidden layers.
python main.py --mode eval --snorm 1 --tnorm 1
@inproceedings{deng2021st,
title={ST-Norm: Spatial and Temporal Normalization for Multi-variate Time Series Forecasting},
author={Deng, Jinliang and Chen, Xiusi and Jiang, Renhe and Song, Xuan and Tsang, Ivor W},
booktitle={Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery \& Data Mining},
pages={269--278},
year={2021}
}
@ARTICLE{9935292,
author={Deng, Jinliang and Chen, Xiusi and Jiang, Renhe and Song, Xuan and Tsang, Ivor W.},
journal={IEEE Transactions on Knowledge and Data Engineering},
title={A Multi-View Multi-Task Learning Framework for Multi-Variate Time Series Forecasting},
year={2022},
volume={},
number={},
pages={1-16},
doi={10.1109/TKDE.2022.3218803}
}