The Drum · Weekly AI news, freeSubscribe →
Safety & Ethics
Federated learning is a privacy-preserving approach to machine learning that trains models on decentralized data without transferring it.
Federated learning allows multiple devices or servers to collaboratively train a model while keeping their data local. Instead of centralizing all the data, each device updates a shared model with its own data, sending only the updated parameters back to a central server. This method ensures that sensitive information remains on the original devices, enhancing privacy and security.
Federated learning is crucial for industries handling sensitive data, such as healthcare and finance. It enables organizations to build robust models without compromising user privacy or violating data protection regulations. For users, it means their personal data can contribute to better services without being exposed to potential breaches.
In federated learning, the process starts with a central server distributing an initial model to participating devices. Each device trains this model using its local data and sends back only the updated parameters. The server aggregates these updates to refine the global model. This cycle repeats until the model reaches optimal performance. By keeping data decentralized, federated learning minimizes exposure to potential security risks.
✗ Federated learning requires all devices to be online simultaneously.
Federated learning can work with devices that are intermittently connected. Devices update the model when they are online, and the central server aggregates these updates over time.
