arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2204.01800cs.DScs.LG
收藏

快速Johnson-Lindenstrauss变换更快

The Fast Johnson-Lindenstrauss Transform is Even Faster

Ora Nova Fandina, Mikael Møller Høgsgaard, Kasper Green Larsen

摘要:

Ailon和Chazelle(SICOMP'09)的开创性快速Johnson-Lindenstrauss(快速JL)变换将$d$维欧几里德空间中的一组$n$点嵌入到最佳$k=O(\varepsilon^{-2} \ln n)$维度中,同时将所有成对距离保持在因子$(1 \pm \varepsilon)$内。快速JL变换支持在$O(d \ln d +k \ln^2 n)$时间内计算数据点的嵌入,其中$d \ln d$项来自与$d \times d$ Hadamard矩阵的乘法,而$k \ln^2 n$项来自与稀疏$k \times d$矩阵的乘法。尽管快速JL变换已经有十多年的历史了,但它是$\varepsilon, d$和$n$之间许多折衷的最快降维技术之一。在这项工作中,我们给出了一个令人惊讶的快速JL变换的新的分析,表明在嵌入时间的$k \ln^2 n$长期可以提高到$(k \ln^2 n)/α$的$α= Ω(\min\{\varepsilon^{-1}\ln(1/\varepsilon), \ln n\})$。通过使用更稀疏的矩阵进行改进。我们还补充了我们改进的分析与下界,表明我们的新分析实际上是紧的。

英文摘要:

The seminal Fast Johnson-Lindenstrauss (Fast JL) transform by Ailon and Chazelle (SICOMP'09) embeds a set of $n$ points in $d$-dimensional Euclidean space into optimal $k=O(\varepsilon^{-2} \ln n)$ dimensions, while preserving all pairwise distances to within a factor $(1 \pm \varepsilon)$. The Fast JL transform supports computing the embedding of a data point in $O(d \ln d +k \ln^2 n)$ time, where the $d \ln d$ term comes from multiplication with a $d \times d$ Hadamard matrix and the $k \ln^2 n$ term comes from multiplication with a sparse $k \times d$ matrix. Despite the Fast JL transform being more than a decade old, it is one of the fastest dimensionality reduction techniques for many tradeoffs between $\varepsilon, d$ and $n$. In this work, we give a surprising new analysis of the Fast JL transform, showing that the $k \ln^2 n$ term in the embedding time can be improved to $(k \ln^2 n)/α$ for an $α= Ω(\min\{\varepsilon^{-1}\ln(1/\varepsilon), \ln n\})$. The improvement follows by using an even sparser matrix. We also complement our improved analysis with a lower bound showing that our new analysis is in fact tight.

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。

↑