Self-hosted XAIN

Documentation Space

Welcome to our documentation space! Learn how to integrate Federated Learning into your AI project with our two Open Source components: the Coordinator and the SDK.

The following overview will guide you through the integration process in easy to replicate steps.

You can check each project's specific documentation here:

Federated Learning

Federated Learning

How it works?

The main protagonists of orchestrating a Federated Learning network are the Coordinator and the Participants.

The Participants are actors in a local environment with access to the data of a single user or database, e.g. from a single device, data bucket or database. Participants, when selected by the Coordinator, train on that user data with an AI model supplied by the Coordinator. This updates the AI model and this update is communicated to the Coordinator.

The Coordinator orchestrates Federated Learning among participants. It initializes the AI model and shares that with participants. Then it selects participants for the next training round, collects updated models from these participants, and aggregates all these updates into a new AI model. This new model is then shared with all participants. This concludes a training round and rounds are repeated iteratively.

XAIN FL Architecture

Quick Guide

How to set up a Federated Learning project

  • 1
    Install dependencies

    Install the following two packages and initalize the SDK: xain-fl and xain-sdk

  • 2
    Register Participants

    Register your participants that you wish to partake in the training rounds. Now you're all set up and ready to train your models with Federated Learning.

  • 3
    Customize metrics

    Optionally, customize the metrics to fit your use case.

You can check each project's specific documentation here:

Here are hello world examples for how to set up your Federated Learning project using either of the very popular AI frameworks provided by Tensorflow or Pytorch.

However, XAIN is framework agnostic and we will publish more examplese in the future.

Talk to us!

If you require further documentation or have feedback on our project, we will look forward to hearing from you.


Unter den Linden 4210117 Berlin, Germany

Copyrights © 2020