Modeling Data Distributions: The Core Idea Behind Generative Models

A generative model is not trying to memorize a dataset. Its real target is the probability structure that could have produced the observed data. That distinction is the foundation of distribution modeling: we can observe finite samples, but we do not know the analytical form of the Data Distribution behind them. Generative modeling therefore...