Comments on: Support vector machines maximizing geometric margins for multi-class classification 3
-
Yoonkyung Lee [1]
-
[1] Ohio State University
Ohio State University
City of Columbus, Estados Unidos
-
- Localización: Top, ISSN-e 1863-8279, ISSN 1134-5764, Vol. 22, Nº. 3, 2014, págs. 852-855
- Idioma: inglés
- DOI: 10.1007/s11750-014-0341-0
- Enlaces
-
Resumen
The paper provides an overview of the multi-objective multi-class support vector machine (MMSVM) based on a series of research articles written by the authors and their collaborators. As an extension of the binary SVM, the MMSVM takes an alltogether approach to classification with multiple categories. Along with the description of the original multi-objective optimization problem for the method, the paper lays out computational strategies for its implementation and further approximation of the solutions to the initial problem via alternative single-objective second-order cone programming problems. The progression from the initial problem to various alternative formulations is methodically presented with summary of the relation between the two sets of corresponding solutions.
