SAT

Training a self-supervised model or download a pretrained self-supervised model.

1. Obtain the pretrained self-supervised model
2. Generating adversarial examples by PGD-KNN
3. Maximize the mutual information between the representations of clean examples and advesarial examples.

AMDIM is selected as the self-supervised model

For detail of the implementation, please refer to the jupyter notebook 'CIFAR_SAT_AMDIM'

@inproceedings{chen2020self,
  title={Self-supervised adversarial training},
  author={Chen, Kejiang and Chen, Yuefeng and Zhou, Hang and Mao, Xiaofeng and Li, Yuhong and He, Yuan and Xue, Hui and Zhang, Weiming and Yu, Nenghai},
  booktitle={ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={2218--2222},
  year={2020},
  organization={IEEE}
}