Theory and implementation of Gaussian Mixture Models and the Expectation Maximization Algorithm.
This .PDF technical report describes the mathematics theory and programming implementation of Gaussian Mixture Models and Expectation Maximization applied to image segmentation. I found myself needing to implement an image segmentation algorithm, but did not have the requisite training data to reproduce state-of-the-art methods in the domain. This motivated me to study the application of Gaussian Mixture Models, a classical machine learning approach that - while unable to produce state-of-the-art results - is fully unsupervised and does not require training data. The accompanying Python reference implementation is available on my GitHub page.