Support vector regression estimation based on non-uniform lost function

Song Xiaofeng, Zhou Tong, Zhang Huanping

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

Abstract

The performances of support vector regression estimation were analyzed. It was found that the insensitive factor ε can affect the performance of support vector regression estimation significantly. The noise inside the sample data should be considered in determining the insensitive factor ε when support vector regression was employed. A novel support vector regression based on non-uniform lost function (NLF-SVR) was proposed to deal with different noise data density function in different region. The formulation and algorithms of computing NLF-SVR were given. The test example showed that the outcomes of NLF-SVR are better than that of conventional SVR. NLF-SVR can be applied in physiological systems modeling.

Original languageEnglish (US)
Title of host publicationProceedings of the 2005 27th Annual International Conference of the Engineering in Medicine and Biology Society, IEEE-EMBS 2005
Pages1127-1130
Number of pages4
StatePublished - Dec 1 2005
Event2005 27th Annual International Conference of the Engineering in Medicine and Biology Society, IEEE-EMBS 2005 - Shanghai, China
Duration: Sep 1 2005Sep 4 2005

Publication series

NameAnnual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings
Volume7 VOLS
ISSN (Print)0589-1019

Other

Other2005 27th Annual International Conference of the Engineering in Medicine and Biology Society, IEEE-EMBS 2005
CountryChina
CityShanghai
Period9/1/059/4/05

Keywords

  • Non-uniform lost function
  • Regression estimator
  • Support vector machine

ASJC Scopus subject areas

  • Signal Processing
  • Biomedical Engineering
  • Computer Vision and Pattern Recognition
  • Health Informatics

Fingerprint Dive into the research topics of 'Support vector regression estimation based on non-uniform lost function'. Together they form a unique fingerprint.

Cite this