Glossary term
Glossary term
Training and Fine-Tuning
Machine learning approach that trains models on decentralised data without centralising raw data, preserving privacy.
A distributed machine learning approach that trains machine learning models using decentralized examples residing on devices such as smartphones. In federated learning, a subset of devices downloads the current model from a central coordinating server. The devices use the examples stored on the devices to make improvements to the model. The devices then upload the model improvements (but not the training examples) to the coordinating server, where they are aggregated with other updates to yield an improved global model. After the aggregation, the model updates computed by devices are no longer needed, and can be discarded.
Since the training examples are never uploaded, federated learning follows the privacy principles of focused data collection and data minimization.
See the Federated Learning comic (yes, a comic) for more details.
Google uses federated learning to improve Gboard's next-word prediction on Android phones - model gradients (not user text) are aggregated from 500M+ devices, improving personalisation without accessing typed content.
NHS England's federated learning project trains diagnostic imaging models across 20 hospital trusts - each trust trains locally on patient CT scans and shares only model gradients, keeping patient data on-site.
Apple uses on-device federated learning to improve autocorrect and Siri without collecting user messages - model updates are computed locally, differentially privatised, and aggregated across millions of devices.
Created for this library
A mobile keyboard team uses federated learning so user typing data improves the next-word model without leaving the device.
A healthcare consortium uses federated learning to train models across hospitals so patient data does not leave each institution.
A bank's research team explores federated learning to combine signals across business units while keeping customer data inside each unit's data perimeter.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License