Previous
Privacy-Efficient Federated Learning for Distributed Medical Intelligence
Medical federated learning must balance privacy, model utility, and communication efficiency. Even when raw images remain local, model or gradient exchange creates privacy attack surfaces, while privacy-transformed or spatially altered inputs can degrade conventional CNN-based learning.
Project Summary
This project investigated lightweight federated learning based on compact Vision Transformer embeddings instead of full parameter or gradient exchange. The design focused on reducing communication cost, limiting privacy exposure, and maintaining useful representation learning under privacy-preserving transformations.