Fedor Pakhurov
RUGitHub ↗
Menu

Course / 02

Generative Models

Probability, latent variables and the main families of modern generative models.

7 sections2 published notes∞ permanent URLs

Available now

Published notes

  1. 01Variational AutoencodersVariational AutoencodersRead →
  2. 02Diffusion ModelsDiffusion ModelsRead →
  1. 01

    Foundations

    • Probability reviewPlanned
    • Maximum likelihoodPlanned
    • Latent variablesPlanned
    • DivergencesPlanned
  2. 02

    Autoregressive Models

    • FactorizationPlanned
    • PixelRNN / PixelCNNPlanned
    • Autoregressive transformersPlanned
  3. 03

    Variational Autoencoders

    • Latent-variable modelsPlanned
    • ELBOPlanned
    • Reparameterization trickPlanned
    • VAE variantsPlanned
  4. 04

    GANs

    • Adversarial trainingPlanned
    • GAN objectivePlanned
    • Training instabilityPlanned
    • GAN variantsPlanned
  5. 05

    Normalizing Flows

    • Change of variablesPlanned
    • Invertible transformationsPlanned
    • Flow architecturesPlanned
  6. 06

    Diffusion Models

    • Forward processPlanned
    • Reverse processPlanned
    • DDPMPlanned
    • Score matchingPlanned
    • SamplingPlanned
    • Latent diffusionPlanned
  7. 07

    Evaluation

    • LikelihoodPlanned
    • FIDPlanned
    • Precision / RecallPlanned
    • Human evaluationPlanned
    • Hallucination analysisPlanned