Flower Enabled Federated Learning

This directory contains data model, code to run federated code for healthcare, and clients and server code.

Project Setup

Start by cloning the project

$ git clone --depth=1 

This will create a new directory ehc_prototype with subfolders and files.

Install dependencies

Run the following command to setup requirements:

cd ehc_prototype
pip install requirements.txt

Note the above requires pip already installed.

Install Mysql

Install mysql on preferred or using OS

Run the SQL script to import database

Connet to mysql database, and then run create a database first.

ehc_prototype;
USE  ehc_prototype;
source /ehc_prototype/sql/ddl.sql

Data

The data it uses is very sensitive, and cannot be publicly shared. however, for you ta work with real data, you can put all data the data/all_data folder. Each along with image files, each image file should be named as filename_label.jpg where label indicates the classified label of the image. the model will automatically extract this information. Moreover, for federated model, split the data as follows:

Two clients Example

Skewed data

Distribute the whole data in >=60 and <=40 splits in data/cleint_1 and data_cleint_2 folders respectively You can play with how you want to order your data to run your models. We chose the ones presented in paper.

Registring Owners and Models

Have a look and run the method in federated/requests/request.py file

Running the federated learning server:

python3 federed/server.py

Running clients - each in a separate terminal.

Before each run in a terminal, set the folder in the client.py and then run

python3 federated/client.py

python3 federated/client.py

Running stand alone locally only

Use the simple.py in the simple folder.

python3 simple/simple.py