Generative Models | Sharif University of Technology Computer Science Generative Models course

This course offers an in-depth exploration of the cutting-edge techniques behind data generation modeling in artificial intelligence. It provides a comprehensive understanding of how generative models can learn complex data distributions and produce novel outputs, such as realistic images, texts, and sounds. To this end, the probabilistic foundations behind both of shallow and deep generative models are introduced. The course equips participants with the theoretical foundations and practical skills necessary to excel in both academic research and industry applications. Participants should be familiar with fundamental concepts in machine learning and deep learning.

Course Calendar

Date	Lecture topics	Course works	Supplementary Materials
1403/7/1	Introduction Slide		A tutorial on DGM
1403/7/3 1403/7/8	Probabilistic graphical models: Representation Directed (Bayesian networks) and undirected models (Markov random fields) slide		Text books[4]-chapter 3 & 4 CS228 notes
1403/7/10 1403/7/15 1403/7/17	Probabilistic graphical models: Exact and approximate inference Variational inference, MCMC sampling slide		Text books[2]-chapter 7.4 Text books[4]-chapter 9 CS228 notes
1403/7/22	Probabilistic graphical models: Learning Learning from complete and incomplete data slide		Text books[4]-chapter 16 & 17 A note on EM CS228 notes
1403/7/24	Causality and causal inference slide from MPI	HW1 Tiny quiz1
1403/7/29 1403/8/1 1403/8/6 1403/8/8	(Deep generative models) Autoregressive Models FVSBN - NADE RNN - Transformer encoder/decoder, Intro. to large language models slide In context learning slide from Princeton-CS	HW2 Tiny quiz2	Text books[2]-chapter 22 harvard-nlp tutorial CSE599i-NADE CSE599i-transformer In context learning paper 1 Transformer intro. Positional encoding intro.
1403/8/13 1403/8/15	Variational Auto-encoder slide		Text books[2]-chapter 21 CSE599i-VAE notes CS236-VAE notes
1403/8/20 1403/8/22	Generative adversarial networks f-Divergences and f-GAN Wasserstein GAN slide	HW3	Text books[2]-chapter 26 CSE599i-GAN notes Goodfellow GAN main paper WGAN main paper WGAN-GP main paper
1403/8/27 1403/8/29	Normalizing flow Generative Flow: NICE, Glow Neural ODE Slide		Text books[2]-chapter 23 Stanford CS236 notes CSE599i-Flow notes Neural ODE paper
1403/9/6 1403/9/11	Energy based models	Mid-term exam	Text books[2]-chapter 24 CSE599i-EBM notes Stanford CS236 EBM1 slides Stanford CS236 EBM2 slides
1403/9/13 1403/9/18 1403/9/20	Score based models Denoising score matching, Sliced score matching Anealed Langevin dynamics, Noise conditional score networks	HW4 Tiny quiz3	Text books[2]-chapter 25 Yang Song blog for score based models CSE599i-score notes Stanford CS236 slides
1403/9/25 1403/9/27	Diffusion models Slide		Text books[2]-chapter 25 DDPM main paper
1403/10/2	Evaluation of Generative Models	HW5 Tiny Quiz4	A paper on evaluation of GMs
1403/10/9	Trustworthy generative ML
	Student presentations Schedule and topics

Course Content

Text books

Related Text books:

• Bishop, Christopher M. and Hugh Bishop, Deep Learning: Foundations and Concepts, Springer
• Murphy, Kevin P, Probabilistic Machine Learning: Advanced Topics, The MIT Press
• Tomczak, Jakub M., Deep Generative Modeling, Springer
• Koller D., Friedman N., Probabilistic Graphical Models, Principles and Techniques, The MIT Press