🏢 Zhejiang Lab

Huber loss minimization ensures accurate and robust mean estimation under user-level differential privacy, especially for imbalanced datasets and heavy-tailed distributions.