Increasing efficiency of SVM by adaptively penalizing outliers

Yiqiang Zhan, Dinggang Shen

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

3 Citations (Scopus)

Abstract

In this paper, a novel training method is proposed to increase the classification efficiency of support vector machine (SVM). The efficiency of the SVM is determined by the number of support vectors, which is usually large for representing a highly convoluted separation hypersurface. We noted that the separation hypersurface is made unnecessarily over-convoluted around extreme outliers, which dominate the objective function of SVM. To suppress the domination from extreme outliers and thus relatively simplify the shape of separation hypersurface, we propose a method of adaptively penalizing the outliers in the objective function. Since our reformulated objective function has the similar format of the standard SVM, the idea of the existing SVM training algorithms is borrowed for training the proposed SVM. Our proposed method has been tested on the UCI machine learning repository, as well as a real clinical problem, i.e., tissue classification in prostate ultrasound images. Experimental results show that our method is able to dramatically increase the classification efficiency of the SVM, without losing its generalization ability.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages539-551
Number of pages13
Volume3757 LNCS
DOIs
Publication statusPublished - 2005 Dec 1
Externally publishedYes
Event5th International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, EMMCVPR 2005 - St. Augustine, FL, United States
Duration: 2005 Nov 92005 Nov 11

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3757 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other5th International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, EMMCVPR 2005
CountryUnited States
CitySt. Augustine, FL
Period05/11/905/11/11

Fingerprint

Outlier
Support vector machines
Support Vector Machine
Hypersurface
Objective function
Extremes
Training Support
Ultrasound Image
Support Vector
Training Algorithm
Domination
Repository
Learning systems
Prostate
Machine Learning
Simplify
Ultrasonics
Tissue
Experimental Results

ASJC Scopus subject areas

  • Computer Science(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Theoretical Computer Science

Cite this

Zhan, Y., & Shen, D. (2005). Increasing efficiency of SVM by adaptively penalizing outliers. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 3757 LNCS, pp. 539-551). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3757 LNCS). https://doi.org/10.1007/11585978_35

Increasing efficiency of SVM by adaptively penalizing outliers. / Zhan, Yiqiang; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 3757 LNCS 2005. p. 539-551 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 3757 LNCS).

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

Zhan, Y & Shen, D 2005, Increasing efficiency of SVM by adaptively penalizing outliers. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 3757 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 3757 LNCS, pp. 539-551, 5th International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, EMMCVPR 2005, St. Augustine, FL, United States, 05/11/9. https://doi.org/10.1007/11585978_35
Zhan Y, Shen D. Increasing efficiency of SVM by adaptively penalizing outliers. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 3757 LNCS. 2005. p. 539-551. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/11585978_35
Zhan, Yiqiang ; Shen, Dinggang. / Increasing efficiency of SVM by adaptively penalizing outliers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 3757 LNCS 2005. pp. 539-551 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{477c788501dd46b2af446f106e13116d,
title = "Increasing efficiency of SVM by adaptively penalizing outliers",
abstract = "In this paper, a novel training method is proposed to increase the classification efficiency of support vector machine (SVM). The efficiency of the SVM is determined by the number of support vectors, which is usually large for representing a highly convoluted separation hypersurface. We noted that the separation hypersurface is made unnecessarily over-convoluted around extreme outliers, which dominate the objective function of SVM. To suppress the domination from extreme outliers and thus relatively simplify the shape of separation hypersurface, we propose a method of adaptively penalizing the outliers in the objective function. Since our reformulated objective function has the similar format of the standard SVM, the idea of the existing SVM training algorithms is borrowed for training the proposed SVM. Our proposed method has been tested on the UCI machine learning repository, as well as a real clinical problem, i.e., tissue classification in prostate ultrasound images. Experimental results show that our method is able to dramatically increase the classification efficiency of the SVM, without losing its generalization ability.",
author = "Yiqiang Zhan and Dinggang Shen",
year = "2005",
month = "12",
day = "1",
doi = "10.1007/11585978_35",
language = "English",
isbn = "3540302875",
volume = "3757 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "539--551",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Increasing efficiency of SVM by adaptively penalizing outliers

AU - Zhan, Yiqiang

AU - Shen, Dinggang

PY - 2005/12/1

Y1 - 2005/12/1

N2 - In this paper, a novel training method is proposed to increase the classification efficiency of support vector machine (SVM). The efficiency of the SVM is determined by the number of support vectors, which is usually large for representing a highly convoluted separation hypersurface. We noted that the separation hypersurface is made unnecessarily over-convoluted around extreme outliers, which dominate the objective function of SVM. To suppress the domination from extreme outliers and thus relatively simplify the shape of separation hypersurface, we propose a method of adaptively penalizing the outliers in the objective function. Since our reformulated objective function has the similar format of the standard SVM, the idea of the existing SVM training algorithms is borrowed for training the proposed SVM. Our proposed method has been tested on the UCI machine learning repository, as well as a real clinical problem, i.e., tissue classification in prostate ultrasound images. Experimental results show that our method is able to dramatically increase the classification efficiency of the SVM, without losing its generalization ability.

AB - In this paper, a novel training method is proposed to increase the classification efficiency of support vector machine (SVM). The efficiency of the SVM is determined by the number of support vectors, which is usually large for representing a highly convoluted separation hypersurface. We noted that the separation hypersurface is made unnecessarily over-convoluted around extreme outliers, which dominate the objective function of SVM. To suppress the domination from extreme outliers and thus relatively simplify the shape of separation hypersurface, we propose a method of adaptively penalizing the outliers in the objective function. Since our reformulated objective function has the similar format of the standard SVM, the idea of the existing SVM training algorithms is borrowed for training the proposed SVM. Our proposed method has been tested on the UCI machine learning repository, as well as a real clinical problem, i.e., tissue classification in prostate ultrasound images. Experimental results show that our method is able to dramatically increase the classification efficiency of the SVM, without losing its generalization ability.

UR - http://www.scopus.com/inward/record.url?scp=33646559874&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33646559874&partnerID=8YFLogxK

U2 - 10.1007/11585978_35

DO - 10.1007/11585978_35

M3 - Conference contribution

AN - SCOPUS:33646559874

SN - 3540302875

SN - 9783540302872

VL - 3757 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 539

EP - 551

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -