Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach (ICLR 2024)

This repository includes codes to demonstrate the integration of DBPM into SimCLR. 📃Read the paper.

Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach \ The Twelfth International Conference on Learning Representations (ICLR 2024) \ Xiang Lan, Hanshu Yan, Shenda Hong, Mengling Feng

Last update on 08 Mar 2024

Dataset and preprocessing

1. We use the PTB-XL dataset for demonstration. Data can be downloaded at PhysioNet PTB-XL.
2. Specify the paths for both raw data and processed data in 'data_processing/process_ptbxl.py'.
3. python process_ptbxl.py

Run task

1. Specify the path for processed data within 'config/config_ecg.yaml'.
2. . run.sh

Main dependencies

python==3.7.10
pytorch==1.11.0
numpy==1.20.3
scikit-learn==0.24.2
scipy==1.6.3

Reference

We appreciate your citations if you find our paper related and useful to your research!

@inproceedings{
lan2024towards,
title={Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach},
author={Xiang Lan and Hanshu Yan and Shenda Hong and Mengling Feng},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=K2c04ulKXn}
}