Measured descent: A new embedding method for finite metrics

R. Krauthgamer, J. R. Lee, M. Mendel, A. Naor

Research output: Contribution to journalArticle

Abstract

We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fréchet embeddings for finite metrics, due to Bourgain (1985) and Rao (1999). We prove that any n-point metric space (X, d) embeds in Hilbert space with distortion O(√αX · log n), where α X is a geometric estimate on the decomposability of X. As an immediate corollary, we obtain an O(√αX · log n) distortion embedding, where λ X is the doubling constant of X. Since λ X ≤ n, this result recovers Bourgain's theorem, but when the metric X is, in a sense, "low-dimensional," improved bounds are achieved. Our embeddings are volume-respecting for subsets of arbitrary size. One consequence is the existence of (k, O(log n)) volume-respecting embeddings for all 1 ≤ k ≤ n, which is the best possible, and answers positively a question posed by U. Feige. Our techniques are also used to answer positively a question of Y. Rabinovich, showing that any weighted n-point planar graph embeds in lO(log n) with O(1) distortion. The O(log n) bound on the dimension is optimal, and improves upon the previously known bound of O((log n)2).

Original languageEnglish (US)
Pages (from-to)839-858
Number of pages20
JournalGeometric and Functional Analysis
Volume15
Issue number4
DOIs
StatePublished - Aug 2005

Fingerprint

Descent
Metric
Metric space
Decomposability
Doubling
Planar graph
Probability Measure
Corollary
Hilbert space
Subset
Arbitrary
Theorem
Estimate

ASJC Scopus subject areas

  • Mathematics(all)
  • Analysis

Cite this

Measured descent : A new embedding method for finite metrics. / Krauthgamer, R.; Lee, J. R.; Mendel, M.; Naor, A.

In: Geometric and Functional Analysis, Vol. 15, No. 4, 08.2005, p. 839-858.

Research output: Contribution to journalArticle

Krauthgamer, R. ; Lee, J. R. ; Mendel, M. ; Naor, A. / Measured descent : A new embedding method for finite metrics. In: Geometric and Functional Analysis. 2005 ; Vol. 15, No. 4. pp. 839-858.
@article{ad930fb18c6343abb3c5e5c2091727be,
title = "Measured descent: A new embedding method for finite metrics",
abstract = "We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fr{\'e}chet embeddings for finite metrics, due to Bourgain (1985) and Rao (1999). We prove that any n-point metric space (X, d) embeds in Hilbert space with distortion O(√αX · log n), where α X is a geometric estimate on the decomposability of X. As an immediate corollary, we obtain an O(√αX · log n) distortion embedding, where λ X is the doubling constant of X. Since λ X ≤ n, this result recovers Bourgain's theorem, but when the metric X is, in a sense, {"}low-dimensional,{"} improved bounds are achieved. Our embeddings are volume-respecting for subsets of arbitrary size. One consequence is the existence of (k, O(log n)) volume-respecting embeddings for all 1 ≤ k ≤ n, which is the best possible, and answers positively a question posed by U. Feige. Our techniques are also used to answer positively a question of Y. Rabinovich, showing that any weighted n-point planar graph embeds in l∞O(log n) with O(1) distortion. The O(log n) bound on the dimension is optimal, and improves upon the previously known bound of O((log n)2).",
author = "R. Krauthgamer and Lee, {J. R.} and M. Mendel and A. Naor",
year = "2005",
month = "8",
doi = "10.1007/s00039-005-0527-6",
language = "English (US)",
volume = "15",
pages = "839--858",
journal = "Geometric and Functional Analysis",
issn = "1016-443X",
publisher = "Birkhauser Verlag Basel",
number = "4",

}

TY - JOUR

T1 - Measured descent

T2 - A new embedding method for finite metrics

AU - Krauthgamer, R.

AU - Lee, J. R.

AU - Mendel, M.

AU - Naor, A.

PY - 2005/8

Y1 - 2005/8

N2 - We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fréchet embeddings for finite metrics, due to Bourgain (1985) and Rao (1999). We prove that any n-point metric space (X, d) embeds in Hilbert space with distortion O(√αX · log n), where α X is a geometric estimate on the decomposability of X. As an immediate corollary, we obtain an O(√αX · log n) distortion embedding, where λ X is the doubling constant of X. Since λ X ≤ n, this result recovers Bourgain's theorem, but when the metric X is, in a sense, "low-dimensional," improved bounds are achieved. Our embeddings are volume-respecting for subsets of arbitrary size. One consequence is the existence of (k, O(log n)) volume-respecting embeddings for all 1 ≤ k ≤ n, which is the best possible, and answers positively a question posed by U. Feige. Our techniques are also used to answer positively a question of Y. Rabinovich, showing that any weighted n-point planar graph embeds in l∞O(log n) with O(1) distortion. The O(log n) bound on the dimension is optimal, and improves upon the previously known bound of O((log n)2).

AB - We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fréchet embeddings for finite metrics, due to Bourgain (1985) and Rao (1999). We prove that any n-point metric space (X, d) embeds in Hilbert space with distortion O(√αX · log n), where α X is a geometric estimate on the decomposability of X. As an immediate corollary, we obtain an O(√αX · log n) distortion embedding, where λ X is the doubling constant of X. Since λ X ≤ n, this result recovers Bourgain's theorem, but when the metric X is, in a sense, "low-dimensional," improved bounds are achieved. Our embeddings are volume-respecting for subsets of arbitrary size. One consequence is the existence of (k, O(log n)) volume-respecting embeddings for all 1 ≤ k ≤ n, which is the best possible, and answers positively a question posed by U. Feige. Our techniques are also used to answer positively a question of Y. Rabinovich, showing that any weighted n-point planar graph embeds in l∞O(log n) with O(1) distortion. The O(log n) bound on the dimension is optimal, and improves upon the previously known bound of O((log n)2).

UR - http://www.scopus.com/inward/record.url?scp=26844509176&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=26844509176&partnerID=8YFLogxK

U2 - 10.1007/s00039-005-0527-6

DO - 10.1007/s00039-005-0527-6

M3 - Article

VL - 15

SP - 839

EP - 858

JO - Geometric and Functional Analysis

JF - Geometric and Functional Analysis

SN - 1016-443X

IS - 4

ER -