GMM with EM – Mr. Joel Kemp

Gaussian Mixture Models (GMM) with Expectation Maximization (EM) is horribly described on 99% of the available material on the internet. Even textbooks make little to no sense (who is the expected audience of these books?) due to the lack of concrete examples and explanations of the theory.

Out of the tons of pages, papers, and tutorials, I have come across a single resource that does a fantastic job of explaining the step-by-step process of segmenting a dataset with GMM and EM.

I’ve implemented the GMM EM algorithm in Matlab and the code can be found in my github portfolio, here