arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2303.01262cs.LGcs.CRcs.ITmath.IT
收藏

基于子集的私有估计实例最优性

Subset-Based Instance Optimality in Private Estimation

Google Research(谷歌研究院)

Travis Dick, Alex Kulesza, Ziteng Sun, Ananda Theertha Suresh

AI总结:

本文提出一种更强的差分隐私估计实例最优性定义,并针对实值数据集构造了在均值、分位数及ℓp-范数最小化器上达到该最优性的私有算法,其中均值算法在多种分布下匹配或超越现有方法。

摘要:

我们为差分隐私估计算法提出了一种新的实例最优性定义。该定义要求最优算法能够同时针对每个数据集$D$,与最佳的私有基准算法竞争,其中基准算法(a)预先知道$D$,且(b)以其在$D$的大子集上的最差性能为评估标准。也就是说,基准算法在向$D$添加可能极端的点时无需表现良好;它只需处理移除少量已存在的真实数据点的情况。这使得我们的基准显著强于先前工作中提出的基准。尽管如此,我们证明,对于实值数据集,在估计广泛类别的数据集属性(包括均值、分位数和$\ell_p$-范数最小化器)时,可以构造实现我们实例最优性概念的私有算法。特别是对于均值,我们提供了详细分析,并表明我们的算法在一系列分布假设下同时匹配或超越现有算法的渐近性能。

英文摘要:

We propose a new definition of instance optimality for differentially private estimation algorithms. Our definition requires an optimal algorithm to compete, simultaneously for every dataset $D$, with the best private benchmark algorithm that (a) knows $D$ in advance and (b) is evaluated by its worst-case performance on large subsets of $D$. That is, the benchmark algorithm need not perform well when potentially extreme points are added to $D$; it only has to handle the removal of a small number of real data points that already exist. This makes our benchmark significantly stronger than those proposed in prior work. We nevertheless show, for real-valued datasets, how to construct private algorithms that achieve our notion of instance optimality when estimating a broad class of dataset properties, including means, quantiles, and $\ell_p$-norm minimizers. For means in particular, we provide a detailed analysis and show that our algorithm simultaneously matches or exceeds the asymptotic performance of existing algorithms under a range of distributional assumptions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑