Learning Kernel Label Decompositions for Ordinal Classification Problems

Research areas:

• Methodology - Machine learning

Year:

2014

Type of Publication:

In Proceedings

Keywords:

Kernel Learning, Support Vector Machines, Ordinal Classification, Kernel-target Alignment

Authors:

• Pérez-Ortiz, María
• Gutiérrez, Pedro Antonio
• Hervás-Martínez, César

Book title:

6th International Conference on Neural Computation Theory and Applications (NCTA2014)

Pages:

218-225

Organization:

Roma (Italy)

Month:

22th-24th October

ISBN:

978-989-758-054-3

BibTex:

@conference{10571861922014, author = "Mar{\'i}a P{\'e}rez-Ortiz and Pedro Antonio Guti{\'e}rrez and C{\'e}sar Herv{\'a}s-Mart{\'i}nez", abstract = "This paper deals with the idea of decomposing ordinal multiclass classification problems when working with kernel methods. The kernel parameters are optimised for each classification subtask in order to better adjust the kernel to the data. More flexible multi-scale Gaussian kernels are considered to increase the goodness of fit of the kernel matrices. Instead of learning independent models for all the subtasks, the optimum convex combination of the kernel matrices is then obtained, leading to a single model able to better discriminate the classes in the feature space. The results of the proposed algorithm shows promising potential for the acquisition of better suited kernels. ", booktitle = "6th International Conference on Neural Computation Theory and Applications (NCTA2014)", isbn = "978-989-758-054-3", keywords = "Kernel Learning, Support Vector Machines, Ordinal Classification, Kernel-target Alignment", month = "22th-24th October", organization = "Roma (Italy)", pages = "218-225", publisher = "SCITEPRESS", title = "{L}earning {K}ernel {L}abel {D}ecompositions for {O}rdinal {C}lassification {P}roblems", url = "www.ijcci.org/Program/2014/Program_Friday.htm", year = "2014", }

Abstract:

This paper deals with the idea of decomposing ordinal multiclass classification problems when working with kernel methods. The kernel parameters are optimised for each classification subtask in order to better adjust the kernel to the data. More flexible multi-scale Gaussian kernels are considered to increase the goodness of fit of the kernel matrices. Instead of learning independent models for all the subtasks, the optimum convex combination of the kernel matrices is then obtained, leading to a single model able to better discriminate the classes in the feature space. The results of the proposed algorithm shows promising potential for the acquisition of better suited kernels.