Active-Passive Federated Learning for Vertically Partitioned Multi-view Data
Researchers have introduced Active-Passive Federated Learning (APFed), a new framework designed to improve the flexibility of vertical federated learning. This approach allows an active client to independently perform model inference after the model is built, even if passive clients become unavailable. The APFed framework has been demonstrated through two classification methods that utilize reconstruction and contrastive losses, showing effective results in experimental tests. AI
IMPACT Enhances privacy-preserving AI by allowing for more robust model inference in distributed systems.