Overview

In recent years, there has been increasing concern about the risks associated with data usage, particularly regarding data sharing. This has led to the introduction of regulations which impose stricter rules on data management. These regulations significantly impact scientific research, especially in the biomedical field, where the sensitivity of the data makes multi-center studies more challenging to conduct.

To address these challenges, federated learning (FL) has gained popularity. FL allows multiple parties to collaboratively train a shared machine learning model using their individual data sources without sharing the data itself, thereby enhancing privacy and security. This is typically realized with the help of a server that receives non-sensitive information from data-holder parties (e.g., parameters from a locally trained model) and aggregates it into a global model.

This course will give an overview of FL concepts, including the operational framework, privacy benefits, and challenges. It will show how FL can be used in bioinformatics, covering both federated versions of established bioinformatics algorithms and federated machine learning algorithms designed for bioinformatics data. In hands-on group exercises, a FL consortium will be simulated using the open-source FL platform framework Flower (https://flower.ai/) and a basic FL algorithm will be developed.

Audience

This course is addressed to life scientists and bioinformaticians, from academia or industry, with an interest in machine learning for bioinformatics applications.

Learning outcomes

At the end of the course, the participants are expected to:

• Develop an understanding of FL concepts, including its operational framework, privacy benefits, and challenges.
• Gain an overview of federated methods in bioinformatics, including federated equivalents of established bioinformatics algorithms as well as federated machine learning algorithms applied to bioinformatics data.
• Acquire hands-on experience using a federated learning framework (Flower).
• Understand the process of developing a federated learning algorithm through hands-on experience in a didactic exercise.

Links

The course https://www.sib.swiss/training/course/20260427_FEDBX

To apply https://www.sib.swiss/training/course-apply/20260427_FEDBX