Federated Learning: Collaborative Machine Learning with a Tutorial on How to Get Started - KDnuggets

Federated learning, also known as collaborative learning, allows training models at scale on data that remains distributed on the devices where they are generated. Sensitive data remains with the owners of said data, where training is conducted, and a centralized training orchestrator of training only sees the contribution of each client through model updates. Federated learning doesn't guarantee privacy on its own (we'll touch on breaking and repairing privacy in federated learning systems later on), but it does make privacy possible. With the public and policy-makers becoming more aware of the data economy, demand for privacy-preserving machine learning is on the rise. As a result, data practices have been garnering increased scrutiny, and research on privacy-respecting tools like federated learning is increasingly active.