Which method trains AI models across devices without sharing raw data?

Which method trains AI models across devices without sharing raw data?

• A. Federated learning
• B. Centralized learning
• C. GDPR (EU)
• D. Personal Information Protection Law (PIPL - China)

Answer: (A)

Federated learning trains AI models across devices by keeping data on each device and only sharing model updates, not raw data. Each device learns from its own data and sends its updates to a central aggregator, which combines them into a global model and sends it back for further local refinement. This approach preserves privacy because raw data never leaves the device, reducing the risk of exposing sensitive information while still enabling the model to learn from diverse data sources. In contrast, centralized learning would require pooling raw data in one place, which conflicts with “without sharing raw data.” Regulations like GDPR or PIPL govern data protection, but they are not methods for training models.

Federated learning trains AI models across devices by keeping data on each device and only sharing model updates, not raw data. Each device learns from its own data and sends its updates to a central aggregator, which combines them into a global model and sends it back for further local refinement. This approach preserves privacy because raw data never leaves the device, reducing the risk of exposing sensitive information while still enabling the model to learn from diverse data sources. In contrast, centralized learning would require pooling raw data in one place, which conflicts with “without sharing raw data.” Regulations like GDPR or PIPL govern data protection, but they are not methods for training models.