SVMMDRBF

Learning a Mahalanobis distance kernel for support vector machine classification

The idea of metric learning and support vector machine (SVM) have learned from each other in many recent work. The classifier obtained from SVM algorithm has many advantages, including largest margin and kernel trick. Meanwhile, metric learning algorithm can get a Mahalanobis distance function which emphasize relevant features and reduce the influence of non-informative features. The combination of these two methods, called as SVM with Mahalanobis distance based radial basis function (MDRBF) kernel, seems to be a good solution for most classification problems. We develop a algorithm which Learns a Mahalanobis distance kernel for support vector machine classification. And this is a demo code. If you find the code is useful, please cite our paper "Jiangyuan Mei, Xianqiang Yang, and Huijun Gao, Learning a Mahalanobis distance kernel for support vector machine classification, Journal of The Franklin Institute, under review."