Generative Models

Generative modeling asks a system to learn enough of a data distribution to create, reconstruct, or assign likelihood to examples. This branch currently covers two complementary families:

• Variational Autoencoders learn a structured latent space through an encoder, decoder, and regularized objective.
• Diffusion Models learn to reverse a gradual noising process.

Stanford CS236 provides a broader course path. Fast-moving model and API developments belong in the Frontier section rather than this conceptual branch.