Federated Learning is a machine learning technique that enables training on a decentralized dataset distributed across multiple devices. Instead of sending data to a central server for processing, the training occurs locally on each device, and only model updates are transmitted to a central server.