Periods – no thanks
You can study this course in your own pace when it best suits you and complete it any time of the year (pardoning lecturer vacations).
Please use piazza (see below) – not e-mail – for course-related communication.
Lectures
Course lectures are available as high-resolution videos on YouTube. As the course is arranged into eight chapters, there is a playlist for each chapter. You can access them from the following links:
- Introduction
- Generative Modeling
- Variational Autoencoders
- Generative Adversarial Networks
- Autoregressive Models
- Normalizing Flow Models
- Energy-Based Models
- Diffusion Models
Slides
There’s a slide set accompanying each chapter. You can download the slides from this link.
Notebooks
Apart from the introductory chapters, there are runnable implementation examples for each. You can run them on your own computer if you have a capable GPU, or – easier but likely slower – on Colab. Course notebooks can be downloaded from this link.
Completing the course
The course is completed in two stages: an oral exam and a subsequent project work. Project work is the main thing – the oral exam is there just to convince me that you have what it takes to do the project.
There are three possible ways of implementing the project (feel free to suggest your own way):
- general theoretical work (e.g. a survey explaining the background of some modeling technique),
- focused theoretical work (detailed review of a significant article in the field), and
- practical work, which familiarizes you with a modern implementation.
While the “practical work” might sound appealing, the computational burden is a difficulty. Don’t select it unless you have access to reasonably powerful computational resources. As to the focused theoretical work, I strongly suggest the “research trends” mentioned in the end of each chapter (with links on the slides). On the other hand, any sufficiently challenging and relevant topic that falls under the umbrella of “deep generative modeling” is acceptable.
The oral exam will consist of at least one question on each course chapter, a general conversation, and your feedback on the course. A well-executed oral exam gives one bonus point in the grading.
Piazza
Join the course piazza site for all discussions and news. Click the link to enroll. Contact me if you have trouble enrolling (e.g. because you don’t have an utu.fi e-mail). If you have something to ask about the course, do it on piazza – don’t mail me. In this way, everyone can see the answers and avoid asking the same thing twice. Everyone is invited to contribute if someone asks something. Please consider posting if you notice some interesting news related to generative modeling, too.
For those who hate to sign in new systems, it is possible to just view the piazza as a restricted, read only version, where all students’ names are anonymized and all student information is hidden.