Xiaochen Li

Federated differentially private data synthesis can now achieve utility comparable to centralized approaches, even with heterogeneous data distributions, thanks to a novel framework that smartly handles noise and redundancy.