Skip to content

xzzjx/dpsgd

Repository files navigation

deep learning with differential privacy using pytorch

Idea

  1. 动态调节sigma和C
  2. 使用stochastic weight average
  3. 使用self-ensemble

Todo

  • 完成fully connected network mnist
  • 完成autoencoder mnist,对mnist做PCA降维处理。保存模型,加载自己数据
  • 完成per example gradient
  • 查看不同层的梯度,以及总梯度的l2norm, l2norm的方差变化的情况
  • 完成privacy accountant
  • 完成dp_pca

Tricks

  1. 使用torch.utils.data.WeightedRandomSampler来实现resampling

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages