Generative Models
Generative models learn the probability distribution of data and can generate realistic new samples. This section covers the mainstream generative modeling paradigms.
Contents:
- VAE — Variational inference, reparameterization trick, ELBO
- GAN — Adversarial training, generator & discriminator, mode collapse
- Diffusion — Forward diffusion, reverse denoising, DDPM & DDIM
- Flow Matching — Continuous normalizing flows, optimal transport, conditional flow matching
- Generative 3D — Diffusion-driven 3D reconstruction and generation, SDS/VSD, multi-view diffusion, LRM/Trellis/Hunyuan3D