Assignments

Assignment 1: Autoregressive Models

out February 27, due March 13, 2026 (23:59:59)

In this homework, you will implement and train a variety of different autoregressive models, mainly PixelCNN and its variants.

Assignment 2: Normalizing Flow Models and VAEs

out March 13, due April 3, 2026 (23:59:59)

In this homework, you will implement and train a normalizing flow model, RealNVP, and various VAE models.

Assignment 3: GANs and Denoising Diffusion Models

out April 3, due April 22, 2026 (23:59:59)

In this homework, you will implement and train CycleGAN and a diffusion-based model.


Late Policy:You may use up to 7 grace days (in total) over the course of the semester. That is, you can submit your solutions without any penalty if you have free grace days left. Any additional unapproved late submission will be punished (1 day late: 20% off, 2 days late: 40% off, 3 days late: 50% off). Note that no submission after 3 days will be accepted -- also means that you may use at most 3 grace days for a specific assignment.