Ethics & Safety Intermediate 2 min read

What is federated learning?

A machine learning approach where models are trained on local data and then shared with a central server to improve overall performance

federated learning explained in plain English

Federated learning is a type of machine learning that allows devices or organizations to collaborate on model training while keeping their data private. This approach enables the creation of more accurate and robust models without requiring the sharing of sensitive data

Analogy

Imagine a group of friends trying to create a map of their neighborhood. Instead of sharing their individual maps, they each create their own and then share the updates with a central mapmaker, who combines the information to create a more accurate and complete map

Example

Google uses federated learning to improve the performance of its virtual keyboard on Android devices. The model is trained on user interactions on individual devices and then shared with a central server to update the overall model

How is federated learning used?

Federated learning is used in applications such as mobile devices, healthcare, and finance, where data privacy is a concern. It allows multiple parties to collaborate on model training while maintaining control over their data

Common misconceptions about federated learning

One common misconception is that federated learning requires the sharing of raw data, which is not the case. Only the trained models or updates are shared, keeping the underlying data private

History

Federated learning was first introduced in 2016 by Google researchers as a way to enable machine learning on mobile devices while preserving user privacy

Also known as

distributed learning collaborative learning privacy-preserving learning