Institute Homepage
Institute Homepage DE Sign In

Back

Empirical Inference Article 2004

Experimentally optimal v in support vector regression for different noise models and parameter settings

Thumb ticker sm l1170153
Empirical Inference
Bernhard Schölkopf
  • Director

In Support Vector (SV) regression, a parameter ν controls the number of Support Vectors and the number of points that come to lie outside of the so-called var epsilon-insensitive tube. For various noise models and SV parameter settings, we experimentally determine the values of ν that lead to the lowest generalization error. We find good agreement with the values that had previously been predicted by a theoretical argument based on the asymptotic efficiency of a simplified model of SV regression. As a side effect of the experiments, valuable information about the generalization behavior of the remaining SVM parameters and their dependencies is gained. The experimental findings are valid even for complex ‘real-world’ data sets. Based on our results on the role of the ν-SVM parameters, we discuss various model selection methods.

Author(s): Chalimourda, A. and Schölkopf, B. and Smola, AJ.
Journal: Neural Networks
Volume: 17
Number (issue): 1
Pages: 127-141
Year: 2004
Month: January
Day: 0
Bibtex Type: Article (article)
DOI: 10.1016/S0893-6080(03)00209-0
Digital: 0
Electronic Archiving: grant_archive
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik
Links:

BibTex 

@article{4680,
  title = {Experimentally optimal v in support vector regression for different noise models and parameter settings},
  journal = {Neural Networks},
  abstract = {In Support Vector (SV) regression, a parameter ν controls the number of Support Vectors and the number of points that come to lie outside of the so-called var epsilon-insensitive tube. For various noise models and SV parameter settings, we experimentally determine the values of ν that lead to the lowest generalization error. We find good agreement with the values that had previously been predicted by a theoretical argument based on the asymptotic efficiency of a simplified model of SV regression. As a side effect of the experiments, valuable information about the generalization behavior of the remaining SVM parameters and their dependencies is gained. The experimental findings are valid even for complex ‘real-world’ data sets. Based on our results on the role of the ν-SVM parameters, we discuss various model selection methods.},
  volume = {17},
  number = {1},
  pages = {127-141},
  organization = {Max-Planck-Gesellschaft},
  school = {Biologische Kybernetik},
  month = jan,
  year = {2004},
  slug = {4680},
  author = {Chalimourda, A. and Sch{\"o}lkopf, B. and Smola, AJ.},
  month_numeric = {1}
}