arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2107.11526cs.LGcs.CRcs.DS
收藏

关于轴向矩形的个体学习的样本复杂性

On the Sample Complexity of Privately Learning Axis-Aligned Rectangles

Ben-Gurion University(本-古里安大学);Google Research(谷歌研究院)

Menachem Sadigurschi, Uri Stemmer

摘要:

我们重新审视学习轴对齐矩形在有限网格$X^d\subseteq{\mathbb{R}}^d$与差分隐私的基本问题。已有的结果表明,该问题的样本复杂度至多为$\min\left\{ d{\cdot}\log|X| \;,\; d^{1.5}{\cdot}\left(\log^*|X| \right)^{1.5}\right\}$ .也就是说,现有的构造要么要求样本复杂度随$\log|X|$线性增长,要么它随维度$d$超线性增长。我们提出了一种新的算法,减少了样本的复杂性,只有$\tilde{O}\left\{d{\cdot}\left(\log^*|X|\right)^{1.5}\right\}$,达到一个维度最佳的依赖,而不需要的样本复杂性增长与$\log|X|$。为了达到这一改进所使用的技术涉及删除“暴露”的数据点在旅途中,在一种时尚的设计,以避免成本的自适应组合定理。这种技术的核心可能是个人的兴趣,引入了一种新的方法来构建高效的私人算法。

英文摘要:

We revisit the fundamental problem of learning Axis-Aligned-Rectangles over a finite grid $X^d\subseteq{\mathbb{R}}^d$ with differential privacy. Existing results show that the sample complexity of this problem is at most $\min\left\{ d{\cdot}\log|X| \;,\; d^{1.5}{\cdot}\left(\log^*|X| \right)^{1.5}\right\}$. That is, existing constructions either require sample complexity that grows linearly with $\log|X|$, or else it grows super linearly with the dimension $d$. We present a novel algorithm that reduces the sample complexity to only $\tilde{O}\left\{d{\cdot}\left(\log^*|X|\right)^{1.5}\right\}$, attaining a dimensionality optimal dependency without requiring the sample complexity to grow with $\log|X|$.The technique used in order to attain this improvement involves the deletion of "exposed" data-points on the go, in a fashion designed to avoid the cost of the adaptive composition theorems. The core of this technique may be of individual interest, introducing a new method for constructing statistically-efficient private algorithms.

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。

↑