Federated Learning (FL) is a distributed approach to machine learning that enables training on a large corpus of decentralized data residing on devices like mobile phones. FL employs the approach of “bringing the code to the data, instead of the data to the code.” Additionally, it addresses the fundamental problems of privacy, ownership, and locality of data.
Here’s a more in depth look at the approach.
There’s a good high-level overview of federated learning on Google’s AI blog. Devices download the current model, improve it by learning using data local to the phone, and then send a small focused model update back to the cloud, where it is averaged with other user updates to improve the shared model. No individual updates are stored in the cloud, and no training data leaves the device.