EEGMatch: Learning with Incomplete Labels for Semi-Supervised EEG-based Cross-Subject Emotion Recognition

• A Pytorch implementation of our under reviewed paper "EEGMatch: Learning with Incomplete Labels for Semi-Supervised EEG-based Cross-Subject Emotion Recognition".
  

• arxiv

  Installation:

• Python 3.7

• Pytorch 1.3.1

• NVIDIA CUDA 9.2

• Numpy 1.20.3

• Scikit-learn 0.23.2

• scipy 1.3.1

Preliminaries

• Prepare dataset: SEED and SEED-IV

Training

• EEGMatch model definition file: model_EEGMatch.py
• Pipeline of the EEGMatch: implementation_EEGMatch.py
• implementation of domain adversarial training: Adversarial_DG.py

  Dataset prepare

• data_prepare_seed.m

  Usage

• After modify setting (path, etc), just run the main function in the implementation_EEGMatch.py

  Acknowledgement

• The implementation code of domain adversarial training is bulit on the dalib code base

Citation

@misc{zhou2023eegmatch, title={EEGMatch: Learning with Incomplete Labels for Semi-Supervised EEG-based Cross-Subject Emotion Recognition}, author={Rushuang Zhou and Weishan Ye and Zhiguo Zhang and Yanyang Luo and Li Zhang and Linling Li and Gan Huang and Yining Dong and Yuan-Ting Zhang and Zhen Liang}, year={2023}, eprint={2304.06496}, archivePrefix={arXiv}, primaryClass={eess.SP} }