English

A Nonlinear Approach to Dimension Reduction

Computational Geometry 2015-06-09 v4 Data Structures and Algorithms Functional Analysis Metric Geometry

Abstract

The l2l_2 flattening lemma of Johnson and Lindenstrauss [JL84] is a powerful tool for dimension reduction. It has been conjectured that the target dimension bounds can be refined and bounded in terms of the intrinsic dimensionality of the data set (for example, the doubling dimension). One such problem was proposed by Lang and Plaut [LP01] (see also [GKL03,MatousekProblems07,ABN08,CGT10]), and is still open. We prove another result in this line of work: The snowflake metric d1/2d^{1/2} of a doubling set Sl2S \subset l_2 embeds with constant distortion into l2Dl_2^D, for dimension DD that depends solely on the doubling constant of the metric. In fact, the distortion can be made arbitrarily close to 1, and the target dimension is polylogarithmic in the doubling constant. Our techniques are robust and extend to the more difficult spaces l1l_1 and ll_\infty, although the dimension bounds here are quantitatively inferior than those for l2l_2.

Keywords

Cite

@article{arxiv.0907.5477,
  title  = {A Nonlinear Approach to Dimension Reduction},
  author = {Lee-Ad Gottlieb and Robert Krauthgamer},
  journal= {arXiv preprint arXiv:0907.5477},
  year   = {2015}
}
R2 v1 2026-06-21T13:31:06.493Z