PrivFusion: A Privacy-preserving Multi-Agent Framework for Harmonizing Distributed Datasets
Researchers have developed PrivFusion, a novel framework designed to harmonize distributed datasets while preserving privacy. This multi-agent system automates the process of aligning semantically similar features across different sites, a crucial step for effective federated learning in sensitive domains like healthcare. Evaluations on COVID-19 datasets show PrivFusion can significantly reduce manual effort and improve data harmonization efficiency. AI
IMPACT Automates data harmonization for federated learning, potentially enabling more robust multi-site AI analytics in sensitive fields.