33
0

Near-Optimal differentially private low-rank trace regression with guaranteed private initialization

Abstract

We study differentially private (DP) estimation of a rank-rr matrix MRd1×d2M \in \mathbb{R}^{d_1\times d_2} under the trace regression model with Gaussian measurement matrices. Theoretically, the sensitivity of non-private spectral initialization is precisely characterized, and the differential-privacy-constrained minimax lower bound for estimating MM under the Schatten-qq norm is established. Methodologically, the paper introduces a computationally efficient algorithm for DP-initialization with a sample size of nO~(r2(d1d2))n \geq \widetilde O (r^2 (d_1\vee d_2)). Under certain regularity conditions, the DP-initialization falls within a local ball surrounding MM. We also propose a differentially private algorithm for estimating MM based on Riemannian optimization (DP-RGrad), which achieves a near-optimal convergence rate with the DP-initialization and sample size of nO~(r(d1+d2))n \geq \widetilde O(r (d_1 + d_2)). Finally, the paper discusses the non-trivial gap between the minimax lower bound and the upper bound of low-rank matrix estimation under the trace regression model. It is shown that the estimator given by DP-RGrad attains the optimal convergence rate in a weaker notion of differential privacy. Our powerful technique for analyzing the sensitivity of initialization requires no eigengap condition between rr non-zero singular values.

View on arXiv
Comments on this paper