Adaptor

Global Adaptive Transformer for Cross-Subject EEG Classification [Paper]

Core idea: cross attention for distribution alignment

This model is somewhat bad, trying the effect of attention on domain adaptation.

Abstract

• We propose a Global Adaptive Transformer, named GAT, for domain adaptation in EEG classification, where cross attention is used to align marginal distributions of source and target domains (subjects).
• Parallel convolution branches are used to capture temporal and spatial features from raw EEG signals.
• We design an adaptive center loss to align the conditional distribution of EEG features.

Requirements:

• Python 3.10
• Pytorch 1.12

Datasets

• BCI_competition_IV2a - acc 76.58% (hold out)
• BCI_competition_IV2b - acc 84.44% (hold out)

Hope this code can be useful. I would appreciate you citing us in your paper. 😊

@article{song2023global,
  title = {Global {{Adaptive Transformer}} for {{Cross-Subject Enhanced EEG Classification}}},
  author = {Song, Yonghao and Zheng, Qingqing and Wang, Qiong and Gao, Xiaorong and Heng, Pheng-Ann},
  year = {2023},
  journal = {IEEE Transactions on Neural Systems and Rehabilitation Engineering},
  volume = {31},
  pages = {2767--2777},
  issn = {1558-0210},
  doi = {10.1109/TNSRE.2023.3285309}
}