项目作者: CS-savvy

项目描述 :
Comparison between federated and centralized training method
高级语言: Python
项目地址: git://github.com/CS-savvy/experiment_with_tf_federated.git
创建时间: 2020-05-07T05:39:34Z
项目社区:https://github.com/CS-savvy/experiment_with_tf_federated

开源协议:

下载


TensorFlow Federated Experiment

This repository is having python scripts to compare the effectiveness of the model trained with federated training methods with Non-federated training methods.

Training

  1. To train model using TF-Federated, you can train_tff.py. You can find the hyperparameters in first few line of code.
  2. use train_keras.py to start training with usual keras non-federated training.
  3. you can also use train_combined.py to train one by one both methods with same dataset and model.

Compare

Run compare_results.py to generate comparison graph for you metrics as it is saved in results folder in there respective sub-folders.

sample plots
Loss Accuracy
loss loss
loss loss