Advances in metric embedding theory

Ittai Abraham*, Yair Bartal, Ofer Neimant

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

52 Scopus citations


Metric Embedding plays an important role in a vast range of application areas such as computer vision, computational biology, machine learning, networking, statistics, and mathematical psychology, to name a few. The theory of metric embedding received much attention in recent years by mathematicians as well as computer scientists and has been applied in many algorithmic applications. A cornerstone of the field is a celebrated theorem of Bourgain which states that every finite metric space on n points embeds in Euclidean space with O(log n) distortion. Bourgain's result is best possible when considering the worst case distortion over all pairs of points in the metric space. Yet, it is possible that an embedding can do much better in terms of the average distortion. Indeed, in most practical applications of metric embedding the main criteria for the quality of an embedding is its average distortion over all pairs. In this paper we provide an embedding with constant average distortion for arbitrary metric spaces, while maintaining the same worst case bound provided by Bourgain's theorem. In fact, our embedding possesses a much stronger property. We define the ℓq -distortion of a uniformly distributed pair of points. Our embedding achieves the best possible ℓq-distortion for all 1 ≤ q ≤ ∞ simultaneously. These results have several algorithmic implications, e.g. an O(l) approximation for the unweighted uncapacitated quadratic assignment problem. The results are based on novel embedding methods which improve on previous methods in another important aspect: the dimension. The dimension of an embedding is of very high importance in particular in applications and much effort has been invested in analyzing it. However, no previous result improved the bound on the dimension which can be derived from Bourgain's embedding. We prove that any metric space on n points embeds into Lp with distortion O(log n) in dimension O(log n). This provides an optimal bound on the dimension of the embedding. Somewhat surprisingly, we show that a further small improvement is possible at a small price in the distortion, obtaining an embedding with distortion O(log 1+θ n) in optimal dimension O(θ-1 log n/log log n), for any θ > 0. It is worth noting that with the small loss in the distortion this improves upon the best known embedding of arbitrary spaces into Euclidean space, where dimension reduction is used. Our techniques also allow to obtain the optimal distortion for embedding into Lp with nearly tight dimension. For any 1 ≤ p ≤ oc and any 1 ≤ k ≤ p, we give an embedding into Lp with distortion O(⌈logn/k⌉) in dimension 20(k) log n. Underlying our results is a novel embedding method. Probabilistic metric decomposition techniques have played a central role in the field of finite metric embedding in recent years. Here we introduce a novel notion of probabilistic metric decompositions which comes particularly natural in the context of embedding. Our new methodology provides a unified approach to all known results on embedding of arbitrary metric spaces. Moreover, as described above, with some additional ideas they allow to get far stronger results. These metric decompositions seem of independent interest.

Original languageAmerican English
Title of host publicationSTOC'06
Subtitle of host publicationProceedings of the 38th Annual ACM Symposium on Theory of Computing
PublisherAssociation for Computing Machinery
Number of pages16
ISBN (Print)1595931341, 9781595931344
StatePublished - 2006
Event38th Annual ACM Symposium on Theory of Computing, STOC'06 - Seattle, WA, United States
Duration: 21 May 200623 May 2006

Publication series

NameProceedings of the Annual ACM Symposium on Theory of Computing
ISSN (Print)0737-8017


Conference38th Annual ACM Symposium on Theory of Computing, STOC'06
Country/TerritoryUnited States
CitySeattle, WA


  • Metric Embedding


Dive into the research topics of 'Advances in metric embedding theory'. Together they form a unique fingerprint.

Cite this