PDF Estimation using Gaussian Mixtures

This section is concerned with the general PDF estimation problem. Let $ p({\bf z})$ be the PDF of $ {\bf z}$ which must be estimated from training samples. If $ p({\bf z})$ is continuous, it may be approximated to arbitrary accuracy by any kernel-based estimator [51], such as the method of Gaussian Mixtures (GM) [52] given enough terms.


Baggenstoss 2017-05-19