File Download
There are no files associated with this item.
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.1145/1734213.1734215
- Scopus: eid_2-s2.0-77952073941
- WOS: WOS:000279362300002
- Find via
Supplementary
- Citations:
- Appears in Collections:
Article: Ultra-low-dimensional embeddings for doubling metrics
Title | Ultra-low-dimensional embeddings for doubling metrics | ||||||
---|---|---|---|---|---|---|---|
Authors | |||||||
Keywords | Dimension reduction Euclidean embedding Metric spaces | ||||||
Issue Date | 2010 | ||||||
Publisher | Association for Computing Machinery, Inc. The Journal's web site is located at http://www.acm.org/jacm | ||||||
Citation | Journal Of The Acm, 2010, v. 57 n. 4, article no. 21 How to Cite? | ||||||
Abstract | We consider the problem of embedding a metric into low-dimensional Euclidean space. The classical theorems of Bourgain, and of Johnson and Lindenstrauss say that any metric on n points embeds into an O(log n)-dimensional Euclidean space with O(log n) distortion. Moreover, a simple "volume" argument shows that this bound is nearly tight: a uniform metric on n points requires nearly logarithmic number of dimensions to embed with logarithmic distortion. It is natural to ask whether such a volume restriction is the only hurdle to low-dimensional embeddings. In other words, do doubling metrics, that do not have large uniform submetrics, and thus no volume hurdles to low dimensional embeddings, embed in low dimensional Euclidean spaces with small distortion In this article, we give a positive answer to this question. We show how to embed any doubling metrics into O(log log n) dimensions with O(log n) distortion. This is the first embedding for doubling metrics into fewer than logarithmic number of dimensions, even allowing for logarithmic distortion. This result is one extreme point of our general trade-off between distortion and dimension: given an n-point metric (V,d) with doubling dimension dim D, and any target dimension T in the range (Ωdim D log log n) ≤ T O ≤ (log n), we show that the metric embeds into Euclidean space ℝ T with O(log n dim D/T) distortion. © 2010 ACM. | ||||||
Persistent Identifier | http://hdl.handle.net/10722/92647 | ||||||
ISSN | 2023 Impact Factor: 2.3 2023 SCImago Journal Rankings: 2.866 | ||||||
ISI Accession Number ID |
Funding Information: The research of T.-H. Hubert Chan was partly supported by the NSF grant CCR-0122581 (The ALADDIN project), the NSF CAREER award CCF-0448095, and by an Alfred P. Sloan Fellowship. The research of A. Gupta was partly supported by the NSF CAREER award CCF-0448095, and by an Alfred P. Sloan Fellowship. | ||||||
References |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Chan, THH | en_HK |
dc.contributor.author | Gupta, A | en_HK |
dc.contributor.author | Talwar, K | en_HK |
dc.date.accessioned | 2010-09-17T10:52:56Z | - |
dc.date.available | 2010-09-17T10:52:56Z | - |
dc.date.issued | 2010 | en_HK |
dc.identifier.citation | Journal Of The Acm, 2010, v. 57 n. 4, article no. 21 | en_HK |
dc.identifier.issn | 0004-5411 | en_HK |
dc.identifier.uri | http://hdl.handle.net/10722/92647 | - |
dc.description.abstract | We consider the problem of embedding a metric into low-dimensional Euclidean space. The classical theorems of Bourgain, and of Johnson and Lindenstrauss say that any metric on n points embeds into an O(log n)-dimensional Euclidean space with O(log n) distortion. Moreover, a simple "volume" argument shows that this bound is nearly tight: a uniform metric on n points requires nearly logarithmic number of dimensions to embed with logarithmic distortion. It is natural to ask whether such a volume restriction is the only hurdle to low-dimensional embeddings. In other words, do doubling metrics, that do not have large uniform submetrics, and thus no volume hurdles to low dimensional embeddings, embed in low dimensional Euclidean spaces with small distortion In this article, we give a positive answer to this question. We show how to embed any doubling metrics into O(log log n) dimensions with O(log n) distortion. This is the first embedding for doubling metrics into fewer than logarithmic number of dimensions, even allowing for logarithmic distortion. This result is one extreme point of our general trade-off between distortion and dimension: given an n-point metric (V,d) with doubling dimension dim D, and any target dimension T in the range (Ωdim D log log n) ≤ T O ≤ (log n), we show that the metric embeds into Euclidean space ℝ T with O(log n dim D/T) distortion. © 2010 ACM. | en_HK |
dc.language | eng | en_HK |
dc.publisher | Association for Computing Machinery, Inc. The Journal's web site is located at http://www.acm.org/jacm | en_HK |
dc.relation.ispartof | Journal of the ACM | en_HK |
dc.subject | Dimension reduction | en_HK |
dc.subject | Euclidean embedding | en_HK |
dc.subject | Metric spaces | en_HK |
dc.title | Ultra-low-dimensional embeddings for doubling metrics | en_HK |
dc.type | Article | en_HK |
dc.identifier.email | Chan, THH:hubert@cs.hku.hk | en_HK |
dc.identifier.authority | Chan, THH=rp01312 | en_HK |
dc.description.nature | link_to_subscribed_fulltext | - |
dc.identifier.doi | 10.1145/1734213.1734215 | en_HK |
dc.identifier.scopus | eid_2-s2.0-77952073941 | en_HK |
dc.identifier.hkuros | 170689 | - |
dc.relation.references | http://www.scopus.com/mlt/select.url?eid=2-s2.0-77952073941&selection=ref&src=s&origin=recordpage | en_HK |
dc.identifier.volume | 57 | en_HK |
dc.identifier.issue | 4 | en_HK |
dc.identifier.spage | article no. 21 | - |
dc.identifier.epage | article no. 21 | - |
dc.identifier.eissn | 1557-735X | - |
dc.identifier.isi | WOS:000279362300002 | - |
dc.publisher.place | United States | en_HK |
dc.identifier.scopusauthorid | Chan, THH=12645073600 | en_HK |
dc.identifier.scopusauthorid | Gupta, A=8354044800 | en_HK |
dc.identifier.scopusauthorid | Talwar, K=8955044700 | en_HK |
dc.identifier.issnl | 0004-5411 | - |