dixiyao/Patch-Shuffling-Transformers

Privacy-Preserving Split Learning via Patch Shuffling over Transformers

Paper: https://ieeexplore.ieee.org/abstract/document/10027647

API of Patch Shuffling

PatchShuffle

function: utilsenc.PatchShuffle(x)->y
x: input feature; y: outputfeature

BatchShuffle

function: utilsenc.BatchPatchPartialShuffle(x,k1)->y
x: input feature; k: proportions of patches not to be shuffle; y: outputfeature

SpectralShuffle

The function is the same as PatchShuffle or BatchShuffle, but first turn models into spectral domain. Please see the example as reference.

Citation Bibtex

@INPROCEEDINGS{patchshuffling,
  author={Yao, Dixi and Xiang, Liyao and Xu, Hengyuan and Ye, Hangyu and Chen, Yingqi},
  booktitle={2022 IEEE International Conference on Data Mining (ICDM)}, 
  title={Privacy-Preserving Split Learning via Patch Shuffling over Transformers}, 
  year={2022},
  pages={638-647},
  doi={10.1109/ICDM54844.2022.00074}
}

D. Yao, L. Xiang, H. Xu, H. Ye and Y. Chen, "Privacy-Preserving Split Learning via Patch Shuffling over Transformers," 2022 IEEE International Conference on Data Mining (ICDM), Orlando, FL, USA, 2022, pp. 638-647, doi: 10.1109/ICDM54844.2022.00074.