CS-savvy/experiment_with_tf_federated

Comparison between federated and centralized training method

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