Distance Metric Learning for Kernel Machines

Recent work in metric learning has significantly improved the state-of-the-art in k-nearest neighbor classification. However, Support vector machines (with RBF kernels) are arguably the most popula