SpecAR-Net

This repository contains the official implementation for the paper SpecAR-Net: Spectrogram Analysis and Representation Network for Time Series (IJCAI-24).

Paper Info

• [SpecAR-Net: Spectrogram Analysis and Representation Network for Time Series], IJCAI 2024, accepted.

• Equal contributors:

  • Yi Dong
  • Liwen Zhang
  • Youcheng Zhang

This repository groups the implemetations of SpecAR-Net and Appendix of the paper.

Requirements

The recommended requirements for SpecAR-Net are specified as follows:

• Python 3.8
• torch==1.13.1
• einops==0.6.0
• matplotlib==3.1.1
• numpy==1.21.6
• pandas==1.3.5
• patool==1.12
• reformer-pytorch==1.4.4
• scikit-learn==1.0.2
• scipy==1.7.3
• sktime==0.17.1
• sympy==1.10.1
• tensorboard==2.11.2
• tqdm==4.65.0

The dependencies can be installed by:

pip install -r requirements.txt

Data

Prepare Data. You can obtain the well pre-processed datasets from

• [[Google Drive]] (https://drive.google.com/file/d/1nrXwdI4kyDUYKBj3ecRDZ2T-azh9i0JU/view?usp=sharing) or
• [[Baidu Drive]] (link: https://pan.baidu.com/s/1Q2BF1ZezEWPK2nmt5ZbUFg?pwd=uh25)

Then place the downloaded data in the folder./dataset. Here is a summary of supported datasets.

Usage

1. Train and evaluate model. We provide the experiment scripts for all benchmarks under the folder ./scripts/. You can reproduce the experiment results as the following examples:

# long-term forecast
bash ./scripts/long_term_forecast/ETT_script/SpecAR_Net_ETTh1.sh
# short-term forecast
bash ./scripts/short_term_forecast/SpecAR_Net_M4.sh
# imputation
bash ./scripts/imputation/ETT_script/SpecAR_Net_ETTh1.sh
# anomaly detection
bash ./scripts/anomaly_detection/PSM/SpecAR_Net.sh
# classification
bash ./scripts/classification/SpecAR_Net.sh

2. Develop your own model.

• Add the model file to the folder ./models. You can follow the ./models/Transformer.py.
• Include the newly added model in the Exp_Basic.model_dict of ./exp/exp_basic.py.
• Create the corresponding scripts under the folder ./scripts.

Citation

If you find this repo useful, please cite our paper.

@inproceedings{SpecAR2024,
  title={SpecAR_Net: Spectrogram Analysis and Representation Network for Time Series},
  author={Yi Dong and Liwen Zhang and Youcheng Zhang and Wen Chen and Shi Peng and Zhe Ma},
  booktitle={International Joint Conference on Artificial Intelligence},
  year={2024},
}

Contact

If you have any questions or suggestions, feel free to contact:

• Yi dong (dongyi0552@126.com)
• Liwen Zhang (lwzhang9161@126.com)
• Youcheng Zhang (youcheng17@163.com)

Or describe it in Issues.

Acknowledgement

We appreciate this github repos: https://github.com/thuml/TimesNet.