Deep Generative Modeling

This lecture introduces deep generative modeling as an unsupervised deep learning approach for learning hidden data structure and generating synthetic examples. It covers latent-variable models, autoencoders, variational autoencoders, generative adversarial networks, latent-space regularization, debiasing, and CycleGAN-based unpaired translation.

Syllabus

- Introduction
- Why care about generative models?
- Latent variable models
- Autoencoders
- Variational autoencoders
- Priors on the latent distribution
- Reparameterization trick
- Latent perturbation and disentanglement
- Debiasing with VAEs
- Generative adversarial networks
- Intuitions behind GANs
- Training GANs
- GANs: Recent advances
- CycleGAN of unpaired translation
- Summary