This is for releasing the source code of the SRDS 2020 paper "End-to-End Evaluation of Federated Learning and Split Learning for Internet of Things".

If you find it is useful and used for publication. Please kindly cite our work as:

@inproceedings{gao2020end,
title={End-to-End Evaluation of Federated Learning and Split Learning for Internet of Things},
author={Gao, Yansong and Kim, Minki and Abuadbba, Sharif and Kim, Yeonjae and Thapa, Chandra and Kim, Kyuyeon and Camtepe, Seyit A and Kim, Hyoungshick and Nepal, Surya},
booktitle={The 39th International Symposium on Reliable Distributed Systems (SRDS)},
year={2020}}

• Experiment video
• Manual for installing pytorch on raspberry pie3
• Paper
• Presentation

This repository contains the implementations of various distributed machine learning models like Federated learning, split learning and ensemble learning

• Python==3.6
• PyTorch==1.5.1

• Python==3.7
• PyTorch==1.0.0

• models directory: has pre-processed training/testing data of MIT arrhythmia ECG database in hdf5 format. If you want, you can upload another preprocessed train and test data here.
• federated_learning directory: source codes of federated learning in ipynb and .py format
• split_learning directory: source codes of split learning in ipynb and .py format
• ensemble_learning directory: source codes of ensemble learning in ipynb and .py format

you need to use ~client.ipynb file

you need to use ~client_rasp.ipynb or ~client_rasp.py file If you run these files, you can see the temperature, memory usage of raspberry pie.

set hyperparameters

• set variable users, in server and client file
• set variable rounds, local_epoch or epochs of training

Running code

• Run the server code first
• After run server, run the clients

input information

• if you run the server, you can see the printed ip address of server
• when you run the client you need the enter order of client and ip address
• if there is no problem, training will be started

Gao Yansong, Kim Minki, Abuadbba Sharif, Kim Yeonjae, Thapa Chandra, Kim Kyuyeon, Camtepe Seyit A, Kim Hyoungshick, Nepal Surya