Linear Regression

Stable Estimators for Fast Private Statistics

Gavin Brown introduces 'insufficient statistics perturbation,' a differentially private linear regression algorithm that achieves optimal sample complexity and speed by employing stable outlier filtering based on statistical leverage.