7. Probabilistic generative modeling

Model how observed data could have been produced using joint, conditional, and latent-variable distributions. Compare likelihood-based, implicit, autoregressive, and energy-based approaches.