Introduced on day 46
Generative model
A model that learns the distribution of the data itself (or of the data given a condition) so that new samples can be drawn from it.
In detail
A discriminative model learns p(y | x) and answers "which label?". A generative model learns p(x), or p(x | c), and answers "what would a typical x look like?". Language models, image generators and synthetic-data tools are generative. Sampling settings such as temperature decide whether draws are typical or varied, and what the model can produce is limited by the data and capacity it was trained with.