A support vector method for multivariate performance measures

DMCA

A support vector method for multivariate performance measures (2005)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Thorsten Joachims

Venue:	Proceedings of the 22nd International Conference on Machine Learning
Citations:	305 - 6 self

BibTeX

@INPROCEEDINGS{Joachims05asupport,
    author = {Thorsten Joachims},
    title = {A support vector method for multivariate performance measures},
    booktitle = {Proceedings of the 22nd International Conference on Machine Learning},
    year = {2005},
    pages = {377--384},
    publisher = {ACM Press}
}

OpenURL

Abstract

This paper presents a Support Vector Method for optimizing multivariate nonlinear performance measures like the F1score. 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. 1.

Keyphrases

support vector method multivariate performance measure multivariate prediction approach multivariate svms multivariate nonlinear performance measure non-linear performance measure particular rocarea polynomial time contingency table large class conventional classification svm special case

DMCA