Federative Synthesis: Generating Data in Federated Settings | AI Knowledge Hub

Generating Data in Federated Settings

Federative synthesis of data is a method for generating synthetic data within a federated learning environment, where data is distributed across multiple parties who cannot or do not want to combine their datasets. This allows the creation of synthetic data that preserves the statistical properties of the distributed data without requiring consolidation.

This approach is particularly crucial for applications where privacy is paramount, such as medicine, finance, and education, where diverse organizations possess valuable data but are restricted by regulations or competition.

Distributed Learning Generator and Discriminator

Gradient exchange instead of data sharing.

Privacy protection through differential privacy.

Generation from a Unified Latent Space

3. Secure Aggregation

Secure aggregation of statistics:

Frequently asked questions