Regression with Label Differential Privacy

Badih Ghazi, Pritish Kamath, Ravi Kumar +4

We study the task of training regression models with the guarantee of label differential privacy (DP). Based on a global prior distribution on label values, which could be obtained…