Welcome to the MEDfl Documentation

MEDfl is a modular and extensible Python framework designed to accelerate research and deployment in Federated Learning (FL), particularly in the medical domain.

MEDfl supports two complementary modes:

• Simulation Mode — a controlled environment with virtual clients that allows researchers to rapidly experiment, test assumptions, and validate hypotheses before moving to real-world deployment.

• Real-World Mode — a production-ready execution layer where FL pipelines run on distributed machines using the MEDfl Agent, Flower, and secure networking. This mode enables seamless deployment across hospitals, institutions, and research teams.

Together, these two modes provide a complete workflow: from idea → experiment → validation → real-world deployment.

For users seeking an even more accessible experience, MEDfl is fully integrated into MEDfl application, a user-friendly desktop application that simplifies the creation and orchestration of FL pipelines. MEDomicsLab bridges the gap between medical researchers and computer scientists, promoting collaboration through intuitive tools, automated workflows, and secure distributed execution.

Explore the sections below to begin working with MEDfl, understand its architecture, or build your own federated learning pipelines.

Getting Started

• Installation
• Documentation

Simulation Mode

MEDfl’s simulation mode enables you to design, analyze, and validate FL pipelines using virtual clients running locally. This environment is ideal for:

• testing hypotheses

• benchmarking FL configurations

• evaluating differential privacy and transfer learning

• exploring architectural variations

• iterating rapidly and safely before real-world deployment

Simulation Tutorials

• Introduction
• Database management
• NetManager
• Transfer Learning
• Learning Manager
• Learning Results
• Model choice
• MEDfl Complete Tutorial (Simulation)

---

Real-World Mode

MEDfl’s real-world mode enables deploying FL pipelines across distributed machines such as hospital servers, institutional workstations, or research nodes. This mode integrates:

• the MEDfl Python Agent

• Flower federated learning protocol

• secure peer-to-peer networking via Tailscale

• environment validation and client orchestration

• robust, production-ready execution

It provides a seamless path from validated simulation to real, distributed collaboration in medical research.

🌍 Real-World Tutorials

• Real-world Federated Learning Tutorial

API Reference

Contents

• MEDfl
  • MEDfl package
  • MEDfl.LearningManager package
  • MEDfl.NetManager package
  • MEDfl.rw package

Indices and Tables

• Index

• Module Index

• Search Page