Support Vector Method for Multivariate Density Estimation Sayan Mukherjee and Vladimir Vapnik

Abstract

A new method for multivariate density estimation is developed based on the Support Vector Machine (SVM) solution of inverse ill-posed problems. This algorithms is consistent and results in a sparse solution. This method compared favorably to both Parzen's method and the Gaussian Mixture Model (GMM) method in experiments. The SVM and Parzen's method are shown to be closely related, result in solutions of similar quality, however the SVM solution is sparse. The SVM and GMM approachs are both sparse but the SVM approach is consistent and in general results in more accurate estimates.