Abstract : We introduce a new support vector machine devoted to the approximation of multi-class discriminant functions. Its training procedure consists in minimizing a new expression of the guaranteed risk. This bound is significantly tighter than the former ones, which should make the implementation of the structural risk minimization inductive principle in the context of multi-class discrimination better grounded.
Yann Guermeur, André Elisseeff, Hélène Paugam-Moisy. A new multi-class SVM based on a uniform convergence result. IEEE International Joint Conference on Neural Networks 2000 - IJCNN 2000, 2000, Come, Italie, pp.183-188. ⟨inria-00099215⟩