This Python package is an open-source tool designed for simulating federated learning and incorporating differential privacy. It empowers researchers and developers to effortlessly create, execute, and assess federated learning pipelines while seamlessly working with various tabular datasets.
The MEDfl package requires python 3.9 or more to be run. If you don't have it installed on your machine, check out the following link Python. It also requires MySQL database.
For now, you can install the MEDflpackage as:
git clone https://github.com/MEDomics-UdeS/MEDfl.git
cd MEDfl
pip install -e .
MEDfl requires a MySQL DB connection, and this is in order to allow users to work with their own tabular datasets, we have created a bash script to install and configure A MySQL DB with phpmyadmin monitoring system, run the following command then change your credential on the MEDfl/scripts/base.py and MEDfl/scripts/db_config.ini files
sudo bash MEDfl/scripts/setup_mysql.sh
Please ensure to modify the base_url parameter in the MEDfl/global_params.yaml file. The base_url represents the path to the MEDfl project on your local machine. Update this value accordingly.
We used sphinx to create the documentation for this project. you can generate and host it locally by compiling the documentation source code using:
cd docs
make clean
make html
Then open it locally using:
cd _build/html
python -m http.server
We have created a complete tutorial for the different functionalities of the package. It can be found here: Tutorial.
MEDfl is an open-source package that welcomes any contribution and feedback. We wish that this package could serve the growing research community in federated learning for health.