References: Teaching machines to imagine

Source material

Source curriculum (structural mirror, cited as further study):
• MIT 6.S191, "Introduction to Deep Learning", Lecture 4: "Deep Generative Modeling"
  Instructors: Alexander Amini and Ava Amini (MIT)
  Course page: https://introtodeeplearning.com
  Code and labs: https://github.com/aamini/introtodeeplearning
  License: MIT (slides, code, and labs); videos are YouTube standard
  Required attribution: "© Alexander Amini and Ava Amini, MIT 6.S191:
    Introduction to Deep Learning, IntroToDeepLearning.com"
Clawdemy's lessons are original prose that follows the pedagogical arc of this
course. We do not reproduce or transcribe the lectures; we cite them as the
recommended companion. Course materials are used under their MIT license with
the attribution above; all rights to the original videos remain with the creators.

Watch this next

MIT 6.S191, Lecture 4: Deep Generative Modeling by Alexander and Ava Amini. The lecture this lesson mirrors. It walks through autoencoders, VAEs, and GANs with the instructors’ diagrams and shows generated samples from each. Pair it with this lesson to see the latent-space morphing and the GAN contest in motion.

Going deeper

A short, durable list. Each link is a specific next step, not a generic pile.

“Generative Adversarial Networks” (Goodfellow et al., 2014). The paper that introduced the GAN and the generator-versus-discriminator contest. The primary source for half of this lesson; the framing of “two models, one capturing the data and one telling real from fake” is right there in the abstract.
“Auto-Encoding Variational Bayes” (Kingma and Welling, 2013). The paper behind the VAE. Denser than the GAN paper, but it is the origin of the “learn a smooth latent space you can sample from” idea this lesson leans on.
The MIT 6.S191 software labs. The generative-modeling lab lets you train a model and generate your own samples, the most direct way to feel the difference between judging data and producing it. MIT-licensed.

Adjacent topics

Where this connects inside the track.

From edges to objects (lesson 5). The previous lesson closed the discriminative vision pair (networks that judge images). This lesson turned the arrow around to networks that produce them, so the two make a natural before-and-after pair.
Generating by denoising: diffusion (lesson 7). The VAE and the GAN are the classic generative designs. The next lesson covers the approach behind many of today’s most striking image generators, which builds an image by starting from pure noise and removing it step by step.