This repository implements all experiments for knowledge transfer in federated learning
Environment: Python 3.8, pytorch
Downloading dependencies
pip3 install -r requirements.txt
To plot the figure, we use the result files from ./results_fig by running CDKT_plot_summary.py
Main File: CDKT_main.py, Setting File: Setting.py, Plot Summary Results: CDKT_plot_summary.py
PATH: Output: ./results_fig, Figures: ./figs, Dataset: ./data
Modify parameters setting in Setting.py to run the code
Please refer the file Tuning results to access different parameter settings of this work.
-- Example:
CDKT + Mnist + RepFull + Subset of Users + Homogeneous Model: RUNNING_ALGS[1], DATASETS[0], Full_model = False, Rep_Full = True , Subset = True, Same_model = True
CDKT + Cifar-10 + Full + Fixed Users + Heterogeneous Model: RUNNING_ALGS[1], DATASETS[2], Full_model = True, Rep_Full = True , Subset = False, Same_model = False
To run the code, use the command in terminal:
python CDKT_main.py
The Results are stored in ./results_fig and figures in ./figs