Title :
Multi-class support vector machines based on the mahalanobis distance
Author :
Wang, Heng-you ; Gao, Yan-fei ; Zhang, Chang-lun
Author_Institution :
Sch. of Sci., Beijing Univ. of Civil Eng. & Archit., Beijing, China
Abstract :
In the last decade, Support vector machine (SVM) has been deeply investigated and it is often used in Hilbert space by the measure of Euclidean distance. In this paper, we present the SVM with mahalanobis distance, and the details of how to compute the mahalanobis distance in the input and the feature space are described. Finally, we apply it to the image classification and compare the results of them. By this, we obtain a sound conclusion.
Keywords :
Hilbert spaces; image classification; support vector machines; Euclidean distance; Hilbert space; image classification; mahalanobis distance; multiclass support vector machines; Accuracy; Euclidean distance; Kernel; Machine learning; Remote sensing; Support vector machines; Training; Image Classification; Mahalanobis Distance; Multi-class; Support Vector Machine; kernel;
Conference_Titel :
Machine Learning and Cybernetics (ICMLC), 2011 International Conference on
Conference_Location :
Guilin
Print_ISBN :
978-1-4577-0305-8
DOI :
10.1109/ICMLC.2011.6016824
