A Quadratic Loss Multi-Class SVM for which a Radius-Margin Bound Applies

Yann Guermeur (1) , Emmanuel Monfrini (1)

1 ABC - Machine Learning and Computational Biology

To set the values of the hyperparameters of a support vector machine (SVM), the method of choice is cross-validation. Several upper bounds on the leave-one-out error of the pattern recognition SVM have been derived. One of the most popular is the radius-margin bound. It applies to the hard margin machine, and, by extension, to the 2-norm SVM. In this article, we introduce the first quadratic loss multi-class SVM: the M-SVM^2. It can be seen as a direct extension of the 2-norm SVM to the multi-class case, which we establish by deriving the corresponding generalized radius-margin bound.

Mots clés: multi-class SVMs; model selection; leave-one-out cross-validation error; radius-margin bounds

Domaines: Apprentissage [cs.LG]

Yann Guermeur, Emmanuel Monfrini. A Quadratic Loss Multi-Class SVM for which a Radius-Margin Bound Applies. Informatica (ISSN 0868-4952) International Journal, 2011, 22 (1), pp.73-96. ⟨hal-00596121⟩