Please use this identifier to cite or link to this item:
http://hdl.handle.net/1942/2867
Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Copelli, M | - |
dc.contributor.author | VAN DEN BROECK, Christian | - |
dc.date.accessioned | 2007-11-20T10:29:48Z | - |
dc.date.available | 2007-11-20T10:29:48Z | - |
dc.date.issued | 2000 | - |
dc.identifier.citation | PHYSICAL REVIEW E, 61(6). p. 6971-6980 | - |
dc.identifier.issn | 1063-651X | - |
dc.identifier.uri | http://hdl.handle.net/1942/2867 | - |
dc.description.abstract | We study a model of unsupervised learning where the real-valued data vectors are isotropically distributed, except for a single symmetry-breaking binary direction B epsilon {- 1, + 1}(N), Onto which the projections have a Gaussian distribution. We show that a candidate vector J undergoing Gibbs learning in this discrete space, approaches the perfect match J = B exponentially. In addition to the second-order "retarded learning" phase transition for unbiased distributions, we show that first-order transitions can also occur. Extending the known result that the center of mass of the Gibbs ensemble has Bayes-optimal performance, we show that taking the sign of the components of this vector (clipping) leads to the vector with optimal performance in the binary space. These upper hounds are shown generally not to be saturated with the technique of transforming the components of a special continuous vector, except in asymptotic limits and in a special linear case. Simulations are presented which are in excellent agreement with the theoretical results. | - |
dc.language.iso | en | - |
dc.publisher | AMERICAN PHYSICAL SOC | - |
dc.title | Unsupervised learning of binary vectors: A Gaussian scenario | - |
dc.type | Journal Contribution | - |
dc.identifier.epage | 6980 | - |
dc.identifier.issue | 6 | - |
dc.identifier.spage | 6971 | - |
dc.identifier.volume | 61 | - |
local.format.pages | 10 | - |
local.bibliographicCitation.jcat | A1 | - |
dc.description.notes | Univ Calif San Diego, Dept Chem & Biochem 0340, La Jolla, CA 92093 USA. Limburgs Univ Ctr, B-3590 Diepenbeek, Belgium.Copelli, M, Univ Calif San Diego, Dept Chem & Biochem 0340, La Jolla, CA 92093 USA. | - |
local.type.refereed | Refereed | - |
local.type.specified | Article | - |
dc.bibliographicCitation.oldjcat | A1 | - |
dc.identifier.doi | 10.1103/PhysRevE.61.6971 | - |
dc.identifier.isi | 000087575400034 | - |
item.accessRights | Closed Access | - |
item.validation | ecoom 2001 | - |
item.contributor | Copelli, M | - |
item.contributor | VAN DEN BROECK, Christian | - |
item.fulltext | No Fulltext | - |
item.fullcitation | Copelli, M & VAN DEN BROECK, Christian (2000) Unsupervised learning of binary vectors: A Gaussian scenario. In: PHYSICAL REVIEW E, 61(6). p. 6971-6980. | - |
Appears in Collections: | Research publications |
SCOPUSTM
Citations
1
checked on Sep 3, 2020
WEB OF SCIENCETM
Citations
1
checked on Jul 18, 2024
Page view(s)
66
checked on Nov 7, 2023
Google ScholarTM
Check
Altmetric
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.