arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1504.06170cs.ITmath.IT
收藏

小宽度,低失真:低复杂度集合的量化随机嵌入

Small Width, Low Distortions: Quantized Random Embeddings of Low-complexity Sets

Laurent Jacques

摘要:

在哪些条件下以及在哪些失真下,我们可以保持低复杂度向量的成对距离,例如,对于结构化集合,如稀疏向量集或低秩矩阵集,当它们映射到有限向量集时?这项工作解决了这个一般性问题,通过具体使用的量化和抖动的随机线性映射相结合,在以下顺序,一个子高斯随机投影在$\mathbb R^M$中的矢量在$\mathbb R^N$,一个随机翻译,或“抖动”,投影矢量和均匀标量量化器的分辨率$δ>0$应用组件。由于这种量化映射,我们首先能够表明,以高概率,嵌入一个有界集$\mathcal K \subset \mathbb R^N$在$δ\mathbb Z^M$可以实现时,在量化和原始域中的距离分别测量与$\ell_1$ -和$\ell_2$ -范数,并提供量化的观察$M$的数量是大的“高斯平均宽度”的平方之前的$\mathcal K$。在这种情况下,我们表明,嵌入实际上是“准等距”,只有遭受的乘法和加法的失真,其幅度减少为$M^{-1/5}$为一般集,和$M^{-1/2}$为结构集,当$M$增加。第二,当人们只对表征映射到相同量化向量的$\mathcal K$的两个元素之间的最大距离感兴趣时,即,的映射的“一致性宽度”,我们表明,对于类似数量的测量和高概率,该宽度衰减为$M^{-1/4}$的一般集和$1/M$的结构时,$M$增加。最后,作为我们工作的一个重要方面,我们还建立了映射的非高斯性如何影响可以嵌入的向量类,或者当$M$增加时,其一致性宽度可证明会衰减。

英文摘要:

Under which conditions and with which distortions can we preserve the pairwise-distances of low-complexity vectors, e.g., for structured sets such as the set of sparse vectors or the one of low-rank matrices, when these are mapped in a finite set of vectors? This work addresses this general question through the specific use of a quantized and dithered random linear mapping which combines, in the following order, a sub-Gaussian random projection in $\mathbb R^M$ of vectors in $\mathbb R^N$, a random translation, or "dither", of the projected vectors and a uniform scalar quantizer of resolution $δ>0$ applied componentwise. Thanks to this quantized mapping we are first able to show that, with high probability, an embedding of a bounded set $\mathcal K \subset \mathbb R^N$ in $δ\mathbb Z^M$ can be achieved when distances in the quantized and in the original domains are measured with the $\ell_1$- and $\ell_2$-norm, respectively, and provided the number of quantized observations $M$ is large before the square of the "Gaussian mean width" of $\mathcal K$. In this case, we show that the embedding is actually "quasi-isometric" and only suffers of both multiplicative and additive distortions whose magnitudes decrease as $M^{-1/5}$ for general sets, and as $M^{-1/2}$ for structured set, when $M$ increases. Second, when one is only interested in characterizing the maximal distance separating two elements of $\mathcal K$ mapped to the same quantized vector, i.e., the "consistency width" of the mapping, we show that for a similar number of measurements and with high probability this width decays as $M^{-1/4}$ for general sets and as $1/M$ for structured ones when $M$ increases. Finally, as an important aspect of our work, we also establish how the non-Gaussianity of the mapping impacts the class of vectors that can be embedded or whose consistency width provably decays when $M$ increases.

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。

↑