Skip to main content

Federated Learning on a Live Network with SyftBox

Welcome to our Federated Learning (FL) tutorial! In this tutorial, we'll create a full-fledged FL workflow using the MNIST handwritten digits dataset. You'll learn to coordinate multiple datasites training neural networks locally, aggregate their model updates, and maintain data privacy throughout the process — all on the live SyftBox network. While we use MNIST for demonstration, the principles and APIs you'll learn apply to any dataset.

Learning Objectives​

By completing this tutorial, you will:

  • Build and deploy a complete FL system, from setting up the aggregator to training models with the MNIST dataset
  • Experience real-world FL by collaborating with other participants on the live SyftBox network
  • Gain hands-on experience keeping data private while training machine learning models in a distributed setting
  • Master the workflow of coordinating multiple participants in a federated learning experiment

Prerequisites​

  • Basic understanding of Python and PyTorch
  • Familiarity with machine learning concepts, especially neural networks
  • Some knowledge of Federated Learning is helpful but not required
  • SyftBox installed on your computer

Expected Time to Complete​

15-30 minutes, depending on your familiarity with the prerequisites.

Tutorial Structure​

🚀 Ready to start your federated learning journey?

Head to the Federated Learning Primer section to learn (or refresh) both the core concepts of federated learning, and how these map to SyftBox's computational flow.