Getting Started

In the following tutorial, you will create a simple project to train models based on the popular MNIST dataset. This dataset is a large collection of handwritten digits and is widely used in the field of Machine learning and image processing. It provides a solid foundation for training and testing Machine learning models, particularly those involved in image recognition tasks.

The tutorial project is written in Python, a versatile and widely-used programming language that is particularly strong in the field of Machine learning. Python's extensive library support and easy-to-read syntax make it an excellent choice for developing Machine learning clients.

The source code on which the tutorial is based for the process of training a neural network is taken from PyTorch's traditional Machine Learning MNIST example. PyTorch is a powerful open-source Machine learning library for Python, and its MNIST example is a well-established benchmark in the field. However, it's important to note that this example does not involve Federated Learning. Hence, you will create it .

To assist you in this process, the DLR Federated Learning Ecosystem provides a small Python package: fl-demonstrator-mnist. This package implements tools and algorithms to train models based on the MNIST database.

It uses the fl-demonstrator-client package as a dependency which facilities all the client-server communication. In the chapter Create your own client you will use this package to create your own federated learning client.

Both packages are open source and can be found on Github.