Kernel domain description with incomplete data

Using instance-specific margins to avoid imputation

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We present a method of performing kernel space domain description of a dataset with incomplete entries without the need for imputation, allowing kernel features of a class of data with missing features to be rigorously described. This addresses the problem that absent data completion is usually required before kernel classifiers, such as support vector domain description (SVDD), can be applied; equally, few existing techniques for incomplete data adequately address the issue of kernel spaces. Our method, which we call instance-specific domain description (ISDD), uses a parametrisation framework to compute minimal kernelised distances between data points with missing features through a series of optimisation runs, allowing evaluation of the kernel distance while avoiding subjective completions of missing data. We compare results of our method against those achieved by SVDD applied to an imputed dataset, using synthetic and experimental datasets where feature absence has a non-trivial structure.We show that our methods can achieve tighter sphere bounds when applied to linear and quadratic kernels. © 2010 IEEE.

Original languageEnglish
Title of host publicationProceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010
Pages2921-2924
Number of pages4
DOIs
Publication statusPublished - 2010
Event2010 20th International Conference on Pattern Recognition - Istanbul, Turkey
Duration: 23 Aug 201026 Aug 2010

Conference

Conference2010 20th International Conference on Pattern Recognition
Abbreviated titleICPR 2010
CountryTurkey
CityIstanbul
Period23/08/1026/08/10

Fingerprint

Classifiers

Keywords

  • Classification, regression, and ranking
  • Feature extraction, reduction, and analysis
  • Support vector machines and kernels

Cite this

Gripton, A., & Lu, W. (2010). Kernel domain description with incomplete data: Using instance-specific margins to avoid imputation. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010 (pp. 2921-2924) https://doi.org/10.1109/ICPR.2010.716
Gripton, Adam ; Lu, Weiping. / Kernel domain description with incomplete data : Using instance-specific margins to avoid imputation. Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. pp. 2921-2924
@inproceedings{53678d906a2c472f8d60db54f2b05434,
title = "Kernel domain description with incomplete data: Using instance-specific margins to avoid imputation",
abstract = "We present a method of performing kernel space domain description of a dataset with incomplete entries without the need for imputation, allowing kernel features of a class of data with missing features to be rigorously described. This addresses the problem that absent data completion is usually required before kernel classifiers, such as support vector domain description (SVDD), can be applied; equally, few existing techniques for incomplete data adequately address the issue of kernel spaces. Our method, which we call instance-specific domain description (ISDD), uses a parametrisation framework to compute minimal kernelised distances between data points with missing features through a series of optimisation runs, allowing evaluation of the kernel distance while avoiding subjective completions of missing data. We compare results of our method against those achieved by SVDD applied to an imputed dataset, using synthetic and experimental datasets where feature absence has a non-trivial structure.We show that our methods can achieve tighter sphere bounds when applied to linear and quadratic kernels. {\circledC} 2010 IEEE.",
keywords = "Classification, regression, and ranking, Feature extraction, reduction, and analysis, Support vector machines and kernels",
author = "Adam Gripton and Weiping Lu",
year = "2010",
doi = "10.1109/ICPR.2010.716",
language = "English",
isbn = "9780769541099",
pages = "2921--2924",
booktitle = "Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010",

}

Gripton, A & Lu, W 2010, Kernel domain description with incomplete data: Using instance-specific margins to avoid imputation. in Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. pp. 2921-2924, 2010 20th International Conference on Pattern Recognition, Istanbul, Turkey, 23/08/10. https://doi.org/10.1109/ICPR.2010.716

Kernel domain description with incomplete data : Using instance-specific margins to avoid imputation. / Gripton, Adam; Lu, Weiping.

Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 2921-2924.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Kernel domain description with incomplete data

T2 - Using instance-specific margins to avoid imputation

AU - Gripton, Adam

AU - Lu, Weiping

PY - 2010

Y1 - 2010

N2 - We present a method of performing kernel space domain description of a dataset with incomplete entries without the need for imputation, allowing kernel features of a class of data with missing features to be rigorously described. This addresses the problem that absent data completion is usually required before kernel classifiers, such as support vector domain description (SVDD), can be applied; equally, few existing techniques for incomplete data adequately address the issue of kernel spaces. Our method, which we call instance-specific domain description (ISDD), uses a parametrisation framework to compute minimal kernelised distances between data points with missing features through a series of optimisation runs, allowing evaluation of the kernel distance while avoiding subjective completions of missing data. We compare results of our method against those achieved by SVDD applied to an imputed dataset, using synthetic and experimental datasets where feature absence has a non-trivial structure.We show that our methods can achieve tighter sphere bounds when applied to linear and quadratic kernels. © 2010 IEEE.

AB - We present a method of performing kernel space domain description of a dataset with incomplete entries without the need for imputation, allowing kernel features of a class of data with missing features to be rigorously described. This addresses the problem that absent data completion is usually required before kernel classifiers, such as support vector domain description (SVDD), can be applied; equally, few existing techniques for incomplete data adequately address the issue of kernel spaces. Our method, which we call instance-specific domain description (ISDD), uses a parametrisation framework to compute minimal kernelised distances between data points with missing features through a series of optimisation runs, allowing evaluation of the kernel distance while avoiding subjective completions of missing data. We compare results of our method against those achieved by SVDD applied to an imputed dataset, using synthetic and experimental datasets where feature absence has a non-trivial structure.We show that our methods can achieve tighter sphere bounds when applied to linear and quadratic kernels. © 2010 IEEE.

KW - Classification, regression, and ranking

KW - Feature extraction, reduction, and analysis

KW - Support vector machines and kernels

UR - http://www.scopus.com/inward/record.url?scp=78149480258&partnerID=8YFLogxK

U2 - 10.1109/ICPR.2010.716

DO - 10.1109/ICPR.2010.716

M3 - Conference contribution

SN - 9780769541099

SP - 2921

EP - 2924

BT - Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010

ER -

Gripton A, Lu W. Kernel domain description with incomplete data: Using instance-specific margins to avoid imputation. In Proceedings - 2010 20th International Conference on Pattern Recognition, ICPR 2010. 2010. p. 2921-2924 https://doi.org/10.1109/ICPR.2010.716