generative-model-families-autoregressive-diffusion-gan-vae

IN premise — entries/2026/06/21/wiki-Deep_learning-chunk-7.md

Created 2026-06-21T09:55:49+00:00

Four main generative model families: autoregressive (GPT-style), diffusion (Stable Diffusion), adversarial (GAN), and variational (VAE).

Dependents

These beliefs depend on this one:

• OUT generative-models-provide-unlimited-training-data — Generative models (GANs, VAEs, diffusion) would provide effectively unlimited synthetic training data — generating privacy-preserving medical images, augmenting scarce datasets, and enabling training without data collection barriers — fundamentally resolving the labeled-data bottleneck that constrains supervised learning.