Peer-reviewed research
Communication-Efficient Learning of Deep Networks from Decentralized Data
H. Brendan McMahan et al., Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS), JMLR Workshop and Conference Proceedings volume 54, 2017
Read the originalRead on October 17, 2026
- What it is
- The paper that introduced the term federated learning. The authors train deep networks on data spread across mobile devices by sending model updates to a server rather than collecting the raw data, and they test the method on five model architectures and four datasets.
- How much weight it carries
- Peer-reviewed at AISTATS 2017 and widely cited. The experiments use benchmark datasets rather than live keyboard or voice traffic, so they show feasibility, not production performance. The arXiv abstract is the open copy used here.
Facts taken from this source
- McMahan and colleagues proposed federated learning, an approach that leaves the training data distributed on mobile devices and learns a shared model by aggregating locally computed updates.
“leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates” (Abstract)