Federated Learning

Executive Definition: Federated Learning — A decentralized machine learning approach where a model is trained across multiple devices or servers holding local data samples, without exchanging the raw data.

Federated learning enables collaborative AI training without centralizing sensitive information. Each participating device or organization trains a local copy of the model on its own data, and only the resulting model updates (gradients or weights) are shared and aggregated. This approach preserves privacy and complies with data localization regulations.

Frequently Asked Questions

How does federated learning protect privacy?

Raw data never leaves its original device or server. Only encrypted model weight updates are transmitted to a central server, making it extremely difficult to reconstruct the original data.

Where is federated learning used today?

It is used by Google for keyboard prediction on smartphones, by healthcare institutions to train diagnostic models without sharing protected patient records, and by financial institutions for collaborative fraud detection.

What are the primary limitations?

Challenges include high communication overhead, handling statistically heterogeneous (non-IID) data across participants, and the need for robust aggregation methods to defend against adversarial model poisoning.