Explore deep generative modeling in this comprehensive lecture from MIT's Introduction to Deep Learning course. Delve into the importance of generative models, latent variable models, and autoencoders. Learn about variational autoencoders, including priors on latent distributions, the reparameterization trick, and applications in debiasing. Discover generative adversarial networks (GANs), their training process, and recent advances like conditional GANs and CycleGANs. Gain insights into the intuitions behind these powerful techniques and their practical applications. Conclude with a brief introduction to diffusion models, preparing you for the cutting edge of generative AI research.
Overview
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
- Conditioning GANs on a specific label
- CycleGAN of unpaired translation
​ - Summary of VAEs and GANs
- Diffusion Model sneak peak
Taught by
https://www.youtube.com/@AAmini/videos
