Density estimation
Unsupervised statistical machine learning

Recall that density estimation is the very essence of classical statistical analysis: we start with a large number of samples x₁,...,x_N and investigate which probability distribution these samples correspond to.

Two types of density estimation methods can be distinguished in the low-dimensional field: parametric and non-parametric.

With parametric methods, we think we know in advance the general structure of the density function p_θ(x) we're looking for, but this function depends on a certain number of parameters θ, and we try to fix the value of these parameters as best we can according to the samples received. For example, we can assume a priori that the variables under consideration follow a Gaussian distribution...