Learning Per-Exemplar Distance Functions == Learning Anisotropic Per-Exemplar Kernels for Non-Parametric Density Estimation When estimating probability densities, one can be parametric or non-parametric. A parametric appro…