Private and polynomial time algorithms for learning Gaussians and beyond

Abstract: We present a fairly general framework for reducing $(\varepsilon, \delta)$ differentially private (DP) statistical estimation to its non-private counterpart. As the main application of this framework, we give a polynomial time and $(\varepsilon,\delta)$-DP algorithm for learning (unrestricted) Gaussian distributions in $\mathbb{R}<sup>d$. The sample complexity of our approach for learning the Gaussian up to total variation distance $\alpha$ is $$\widetilde{O}(d^2/\alpha² + d^{2\sqrt{\ln(1/\delta)}/\alpha} \varepsilon + d\ln(1/\delta) / \alpha \varepsilon)$$ matching (up to logarithmic factors) the best known information-theoretic (non-efficient) sample complexity upper bound due to Aden-Ali, Ashtiani, and Kamath (ALT'21). In an independent work, Kamath, Mouzakis, Singhal, Steinke, and Ullman (arXiv:2111.04609) proved a similar result using a different approach and with $O(d<sup>{5/2})$ sample complexity dependence on $d$. As another application of our framework, we provide the first polynomial time $(\varepsilon, \delta)$-DP algorithm for robust learning of (unrestricted) Gaussians with sample complexity $\widetilde{O}(d<sup>{3.5})$. In another independent work, Kothari, Manurangsi, and Velingker (arXiv:2112.03548) also provided a polynomial time $(\varepsilon, \delta)$-DP algorithm for robust learning of Gaussians with sample complexity $\widetilde{O}(d<sup>8)$.