Federated Learning (Glossary)

Federated Learning is a term from the fields of Artificial Intelligence, Big Data and Smart Data, as well as Cybersecurity. It is a special method with which many computers work together to train artificial intelligence – without them having to send their own data to a central location.

Imagine many hospitals want to jointly develop an AI that can detect diseases in X-ray images. Instead of sending all sensitive patient data to a central system, each dataset remains within its respective hospital. The AI is trained on-site, and only the learned knowledge – the improved computational rules – is anonymised and merged into a common model. This way, everyone benefits from each other without anyone having to release their sensitive data.

Federated Learning therefore protects private data and meets important data protection requirements, for example, under GDPR. This method is increasingly being used, for instance, in healthcare, on smartphones, or in connected cars. This allows companies to work together on innovations without revealing their data.