Hierarchical Generative Model

Hierarchical Generative Model is a Generative Model that represents data at multiple levels of abstraction. Higher levels capture coarse, global structure and lower levels capture fine detail, with each level conditioned on the one above.

Generating Diverse High-Fidelity Images with VQ-VAE-2 is an example: it uses a hierarchy of top and bottom latent codes to generate high-fidelity images.

See also Hierarchical Model.