Track

Generative Models

Goal: Learn how models sample new data: autoregressive likelihoods, VAEs, GANs, normalizing flows, then diffusion and flow matching.

Prereqs: Deep Learning. Computer Vision helps for the image papers.

Status: done

Work through the steps in order. Bold links open YouTube.

Step	Concept	YouTube	Read
1	Generative modeling landscape		Deep Learning Book — Deep generative models (ch. 20)
2	Autoregressive PixelRNN and PixelCNN	Hugo Larochelle — Autoregressive Generative Models with Deep Learning	van den Oord et al. 2016 — Pixel Recurrent Neural Networks
3	Variational autoencoders	Variational Autoencoder (VAE) from scratch	Kingma & Welling 2013 — Auto-Encoding Variational Bayes
4	Generative adversarial networks	Yannic Kilcher — Generative Adversarial Networks (Classic)	Goodfellow et al. 2014 — Generative Adversarial Networks
5	Style-based GAN generators	A Style-Based Generator Architecture for Generative Adversarial Networks	Karras et al. 2019 — A Style-Based Generator Architecture for Generative Adversarial Networks
6	Normalizing flows	What are Normalizing Flows?	Lilian Weng — Flow-based Deep Generative Models
7	Diffusion models	AI Coffee Break — Diffusion models explained	Lilian Weng — What are Diffusion Models?
8	Denoising diffusion probabilistic models	Outlier — Diffusion Models \| Paper Explanation \| Math Explained	Ho et al. 2020 — Denoising Diffusion Probabilistic Models
9	Score-based generative modeling	Score-Based Generative Modeling through Stochastic Differential Equations	Song et al. 2021 — Score-Based Generative Modeling through Stochastic Differential Equations
10	Classifier-free guidance		Ho & Salimans 2022 — Classifier-Free Diffusion Guidance
11	Latent diffusion	How does Stable Diffusion work? — Latent Diffusion Models Explained	Rombach et al. 2022 — High-Resolution Image Synthesis with Latent Diffusion Models
12	Flow matching		Lipman et al. 2023 — Flow Matching for Generative Modeling

Found a broken link or an unclear step? Report a problem with this track.