ML Anthology

A Support Vector Method for Multivariate Performance Measures

Thorsten Joachims
ICML 2005 pp. 377-384
doi:10.1145/1102351.1102399 /icml/2005/joachims2005icml-support/

Abstract

This paper presents a Support Vector Method for optimizing multivariate nonlinear performance measures like the F1-score. Taking a multivariate prediction approach, we give an algorithm with which such multivariate SVMs can be trained in polynomial time for large classes of potentially non-linear performance measures, in particular ROCArea and all measures that can be computed from the contingency table. The conventional classification SVM arises as a special case of our method.

PDF ICML Semantic Scholar

Cite

Text

Joachims. "A Support Vector Method for Multivariate Performance Measures." International Conference on Machine Learning, 2005. doi:10.1145/1102351.1102399

Markdown

[Joachims. "A Support Vector Method for Multivariate Performance Measures." International Conference on Machine Learning, 2005.](https://mlanthology.org/icml/2005/joachims2005icml-support/) doi:10.1145/1102351.1102399

BibTeX

@inproceedings{joachims2005icml-support,
  title     = {{A Support Vector Method for Multivariate Performance Measures}},
  author    = {Joachims, Thorsten},
  booktitle = {International Conference on Machine Learning},
  year      = {2005},
  pages     = {377-384},
  doi       = {10.1145/1102351.1102399},
  url       = {https://mlanthology.org/icml/2005/joachims2005icml-support/}
}