Outlier detection using ball descriptions with adjustable metric

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

Outlier detection using ball descriptions with adjustable metric

Conference Paper (2006)

Author(s)

DMJ Tax (TU Delft - Electrical Engineering, Mathematics and Computer Science)

P Juszczak (TU Delft - Electrical Engineering, Mathematics and Computer Science)

EM Pekalska (TU Delft - Electrical Engineering, Mathematics and Computer Science)

RPW Duin (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group

Multimedia Computing

CWTS 0.75 <= JFIS < 2.00 Techniek Conference contrib. refereed Wiskunde en Informatica Technische Wiskunde en Informatica

To reference this document use

https://resolver.tudelft.nl/uuid:aed5632a-3036-481e-812c-319d49c495f8

More Info

expand_more

Publication Year

2006

Research Group

Multimedia Computing

Pages (from-to)

587-595

Publisher

Springer

ISBN (print)

3-540-37236-9

Event

Joint IAPR International Workshops SSPR 2006 and SPR 2006, Hong Kong, China (2006-08-17 - 2006-08-19), Heidelberg

Downloads counter

134

Abstract

Sometimesnoveloroutlierdatahastobedetected.Theoutliersmayindicatesomeinterestingrareevent,ortheyshouldbedisregardedbecausetheycannotbereliablyprocessedfurther.Intheidealcasethattheobjectsarerepresentedbyverygoodfeatures,thegenuinedataformsacompactclusterandagoodoutliermeasureisthedistancetotheclustercenter.Thispaperproposesthreenewformulationsto¿ndagoodclustercentertogetherwithanoptimizedp-distancemeasure.Experimentsshowthatforsomerealworlddatasetsverygoodclassi¿cationresultsareobtainedandthat,morespeci¿cally,the1-distanceisparticularlysuitedfordatasetscontainingdiscretefeaturevalues.

No files available

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