In this paper, we focus on differentially private point and interval estimators for simple linear regression. Motivated by recent work that highlights the strong empirical performance of a robust algorithm called $\texttt{DPTheilSen}$, we provide a theoretical analysis of its privacy and accuracy guarantees, offer guidance on setting hyperparameters, and show how to produce differentially private confidence intervals for the slope.
翻译:在本文中,我们侧重于简单的线性回归的有区别的私人点和间距估计值。 受到最近的工作的激励,它突显了称为$\ textt{DPTheilSen}$的强健算法的强力实证表现,我们提供了对其隐私和准确性保障的理论分析,为设置超参数提供了指导,并展示了如何为斜坡生成差异性私人信任间隔。