TsingZ0 / PFLlib

37 traditional FL (tFL) or personalized FL (pFL) algorithms, 3 scenarios, and 20 datasets.
GNU General Public License v2.0
1.35k stars 283 forks source link

关于实验结果 #193

Closed Aaron2Woo closed 2 months ago

Aaron2Woo commented 2 months ago

感谢分享!跑出来的FedAvg@Cifar10的实验结果如下图,存在两个疑问:

  1. 为什么在0-500次round精度均低于10%?
  2. 为什么需要设置这么多的communication round才能收敛?

123

TsingZ0 commented 2 months ago

具体的训练情况与具体场景超参数设置等都相关,并不会有一个稳定统一的表现,对于每个算法来说都是这样