ClusterFL

This is the repo for MobiSys 2021 paper: " ClusterFL: A Similarity-Aware Federated Learning System for Human Activity Recognition " and TOSN 2022 paper: " ClusterFL: A Clustering-based Federated Learning System for Human Activity Recognition ".

Requirements

The program has been tested in the following environment:

Ubuntu 18.04
Python 3.6.8
Tensorflow 2.4.0
sklearn 0.23.1
opencv-python 4.2.0
Keras-python 2.3.1
numpy 1.19.5

ClusterFL Overview

ClusterFL on client:
- do local training with the collabrative learning variables;
- communicate with the server.
ClusterFL on server:
- recieve model weights from the clients;
- learn the relationship of clients;
- update the collabrative learning variables and send them to each client.

Project Strcuture

|-- client                    // code in client side
    |-- client_cfmtl.py/	// main file of client 
    |-- communication.py/	// set up communication with server
    |-- data_pre.py/		// prepare for the FL data
    |-- model_alex_full.py/ 	// model on client 
    |-- desk_run_test.sh/	// run client 

|-- server/    // code in server side
    |-- server_cfmtl.py/        // main file of client
    |-- server_model_alex_full.py/ // model on server 

|-- README.md

|-- pictures               // figures used this README.md

Quick Start

Download the dataset folders (collected by ourself) from FL-Datasets-for-HAR to your client machine.
Chooose one dataset from the above four datasets and change the "read-path" in 'data_pre.py' to the path on your client machine.
Change the 'server_x_test.txt' and 'server_y_test.txt' according to your chosen dataset, default as the one for "imu_data_7".
Change the "server_addr" and "server_port" in 'client_cfmtl.py' as your true server address.
Run the following code on the client machine
```
cd client
./desk_run_test.sh
```
Run the following code on the server machine
```
cd server
python3 server_cfmtl.py
```

Citation

The code and datasets of this project are made available for non-commercial, academic research only. If you would like to use the code or datasets of this project, please cite the following papers:

@inproceedings{ouyang2021clusterfl,
  title={ClusterFL: a similarity-aware federated learning system for human activity recognition},
  author={Ouyang, Xiaomin and Xie, Zhiyuan and Zhou, Jiayu and Huang, Jianwei and Xing, Guoliang},
  booktitle={Proceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Services},
  pages={54--66},
  year={2021}
}