Source: Google Research
Published: April 2017
Circulated: May 28, 2020
Federated learning is a machine learning (ML) technique that enables edge devices (e.g., mobile phones) to collaboratively learn a shared prediction model.
How is this different than traditional ML?
Standard ML approaches require centralizing the training data in a datacenter (i.e., the cloud). Federated learning keeps all of your training data on device.
What are the benefits?
Users get increased data privacy, lower latency, and less power consumption for their devices.
How does it work?
A user’s device downloads the current ML model from a server.
The model is trained on user data stored on their phone.
The model is improved. Revisions are sent to the server.
The server averages the revisions across many users and applies updates to the model.
What’s an example?