Non-Euclidean or non-metric measures can be informative

None, None; None, None; None, None; None, None; None, None

Non-Euclidean or non-metric measures can be informative

Conference Paper (2006)

Author(s)

EM Pekalska (TU Delft - Multimedia Computing)

A Harol (TU Delft - Multimedia Computing)

Robert P.W. Duin (TU Delft - Multimedia Computing)

B Spillmann (External organisation)

H Bunke (External organisation)

Multimedia Computing

CWTS 0.75 <= JFIS < 2.00 Conference contrib. refereed

To reference this document use:

https://resolver.tudelft.nl/uuid:c745ce43-53ad-4848-80e0-8bf9e04b1bbb

More Info

expand_more

Publication Year

2006

Multimedia Computing

Pages (from-to)

871-880

ISBN (print)

3-540-37236-9

Abstract

StatisticallearningalgorithmsoftenrelyontheEuclideandistance.Inpractice,non-Euclideanornon-metricdissimilaritymeasuresmayarisewhencontours,spectraorshapesarecomparedbyeditdistancesorasaconsequenceofrobustobjectmatching[1,2].Itisanopenissuewhethersuchmeasuresareadvantageousforstatisticallearningorwhethertheyshouldbeconstrainedtoobeythemetricaxioms.
Thek-nearestneighbor(NN)ruleiswidelyappliedtogeneraldissimilaritydataasthemostnaturalapproach.Alternativemethodsexistthatembedsuchdataintosuitablerepresentationspacesinwhichstatisticalclassi¿ersareconstructed[3].Inthispaper,weinvestigatetherelationbetweennon-Euclideanaspectsofdissimilaritydataandtheclassi¿cationperformanceofthedirectNNruleandsomeclassi¿erstrainedinrepresentationspaces.Thisisevaluatedonaparameterizedfamilyofeditdistances,inwhichparametervaluescontrolthestrengthofnon-Euclideanbehavior.Our¿ndingisthatthediscriminativepowerofthismeasureincreaseswithincreasingnon-Euclideanandnon-metricaspectsuntilacertainoptimumisreached.Theconclusionisthatstatisticalclassi¿ersperformwellandtheoptimalvaluesoftheparameterscharacterizeanon-Euclideanandsomewhatnon-metricmeasure

No files available

Metadata only record. There are no files for this record.