Credit: Scientific Reports, July 2020
Federated learning -- an approach first implemented by Google for keyboards' autocorrect functionality -- trains an algorithm across multiple decentralized devices or servers holding local data samples, without exchanging them.