wenzhu23333 / Differential-Privacy-Based-Federated-Learning

Everything you want about DP-Based Federated Learning, including Papers and Code. (Mechanism: Laplace or Gaussian, Dataset: femnist, shakespeare, mnist, cifar-10 and fashion-mnist. )
GNU General Public License v3.0
348 stars 56 forks source link

想问下关于Sensitivity计算的问题 #8

Closed xiyuanyang45 closed 1 year ago

xiyuanyang45 commented 1 year ago

我看这个MA的sensitivity值计算的函数 def cal_sensitivity_MA(lr, clip, dataset_size): return lr * clip / dataset_size 想问下为什么还要乘一个learning rate呢😭实在是想不明白

wenzhu23333 commented 1 year ago

因为目前是对参数加噪,不是对梯度加躁。 参考:https://yangwenzhuo.top/2022/05/03/%E8%81%94%E9%82%A6%E5%AD%A6%E4%B9%A0%EF%BC%88%E5%9B%9B%EF%BC%89/

xiyuanyang45 commented 1 year ago

懂了,谢谢!