Differentially Private Learning with Margin Guarantees

Neural Information Processing Systems 

We present a series of new differentially private (DP) algorithms with dimension-independent margin guarantees.