Robust 1-norm soft margin smooth support vector machine

Li Jen Chien*, Yuh-Jye Lee, Zhi Peng Kao, Chih Cheng Chang

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

Based on studies and experiments on the loss term of SVMs, we argue that 1-norm measurement is better than 2-norm measurement for outlier resistance. Thus, we modify the previous 2-norm soft margin smooth support vector machine (SSVM 2) to propose a new 1-norm soft margin smooth support vector machine (SSVM 1). Both SSVMs can be solved in primal form without a sophisticated optimization solver. We also propose a heuristic method for outlier filtering which costs little in training process and improves the ability of outlier resistance a lot. The experimental results show that SSVM 1 with outlier filtering heuristic performs well not only on the clean, but also the polluted synthetic and benchmark UCI datasets.

Original languageEnglish
Title of host publicationIntelligent Data Engineering and Automated Learning, IDEAL 2010 - 11th International Conference, Proceedings
Pages145-152
Number of pages8
DOIs
StatePublished - 8 Nov 2010
Event11th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2010 - Paisley, United Kingdom
Duration: 1 Sep 20103 Sep 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6283 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference11th International Conference on Intelligent Data Engineering and Automated Learning, IDEAL 2010
CountryUnited Kingdom
CityPaisley
Period1/09/103/09/10

Keywords

  • Classification
  • Outlier resistance
  • Robustness
  • Smooth technique
  • Support vector machine

Fingerprint Dive into the research topics of 'Robust 1-norm soft margin smooth support vector machine'. Together they form a unique fingerprint.

Cite this