On: Tale of Three Probabilistic Families: Discriminative, Descriptive and Generative

[rsp1984]

The descriptive model specifies the probability distribution of the signal, based on an energy function defined on the signal.

I understand Discriminative and Generative models and their usage in inference of a quantity from a given signal.

However I don't quite get what a Descriptive model does. What is it used for? It doesn't sound like inference is the goal.