Federated Learning

Federated learning trains models across decentralized devices or silos by sharing updates rather than raw data.

Devices learn locally and send model updates—not their private files—to build a shared model.

Clients train locally and share model updates for aggregation. Challenges include non-IID data, communication cost, and privacy of updates.

Federated learning matters where data cannot centrally pool—mobile, healthcare, and multi-institution settings.

• On-device personalization
• Cross-silo enterprise learning
• Privacy-sensitive collaboration

• Keeps raw data local
• Enables collaboration across silos

• Harder optimization
• Updates can still leak information without extra protections

• Federated ≠ automatically differentially private

Is federated learning widely deployed?

Yes in some consumer and research settings; enterprise adoption varies with infrastructure maturity.

• NIST AI RMF

• McMahan et al. — Communication-Efficient Learning of Deep Networks from Decentralized Data