Deep Generative Models / Fall 2024
Updates
-
Late Homework Policy. Each student will be given three (3) late days in total for this course. You can freely arrange them across all homework. For example, you can use two days for HW1, one day for HW2, and you will have to submit HW3 exactly on time. Another example is that, if you submit HW1 and HW2 on time, you will have three days for HW3.
- New Lecture is up: Linear Dynamical Systems [slides] [recording 1] [recording 2]
- New Lecture is up: Hidden Markov Models [slides] [recording 1] [recording 2]
Course Description
Generative models have found widespread applications in science and engineering. Recent progress in deep learning has enabled the application of generative models to complex high-dimensional data such as images, videos, text and speech. This course will cover state-of-the-art deep generative models, including variational autoencoders (VAEs), auto-regressive models, diffusion models, and generative adversarial networks (GANs). The course will also illustrate various applications of deep generative models to image and video generation, text and speech generation, image captioning, text-to-image generation, and inverse problems.
Our course on Canvas.
