Conceptual
Login

Density Estimation with Mixture Models

A fitted GMM is a full generative density: it assigns a likelihood to any point (enabling anomaly scoring via low log-density), can be sampled from by drawing a component then a Gaussian draw, and marginals/conditionals stay analytically Gaussian-mixture.

This Concept is waiting for its first lesson!

A fitted GMM is a full generative density: it assigns a likelihood to any point (enabling anomaly scoring via low log-density), can be sampled from by drawing a component then a Gaussian draw, and marginals/conditionals stay analytically Gaussian-mixture.

Are you a teacher? Sign in to start contributing.

Sign In