What Is Gaussian Mixture Model (GMM)?

Gaussian mixture models are, in a word, a-maz-ing. As we all know, they're a great way to model non-Gaussian data and are fantastic at approximating the shape of the distribution of random vector elements. What do you do when you want to add more than one Gaussian? Well, that's where things get interesting. Let's say you want to add three Gaussians. You'd take one of them and add it to your mixture - that would be your first component density. Then you'd add two more together - that would be your second component density. The third component density will be another copy of what you said last time! A mixture is created Model with three components by adding three elements (M=3). In mathematics terms, this is denoted as p(x|λ) = X M i=1 wi g(x|µi, Σi), where M is denoted for mixture weights, x is the continuous-valued data vector from the D-dimension and g(x|µi, Σi ) is the Gaussian component densities Gaussian mixture models are like the Bieber of clustering, density estimation, and classification methods: They're just so hot right now. You see, Gaussian mixture models are based on the assumption that mixtures of Gaussian distributions can describe data points in a given set. This means they're good at figuring out what you look like based on your voice or face or distinguishing between different types of objects. They're also super-easy to use! So if you're looking for a way to describe data without getting into the weeds too much, then you might want to try using this Model. For example: if you're going to take pictures from different angles and get an idea of how tall an object is, then a Gaussian mixture model might be just what you need.