Computer Science 690L - Deep Generative Models

This course offers an introduction to the probabilistic foundations and learning algorithms of generative models, with a focus on deep learning architectures. We will delve into generative models as conditional probability distributions represented by p(x|y), where x is a high-dimensional random vector, and y can be either high or low-dimensional. The curriculum encompasses various facets of generative models, including sampling, density estimation, training techniques, the exploration of latent spaces, and architectural considerations. We'll also examine the practical applications of these models in tasks such as data generation, imputation, and latent space interpolation. Accompanying the course are hands-on projects and exercises that allow students to investigate the scalability of various methods in real-world scenarios, as well as their theoretical underpinnings.

Open to Masters and PhD Computer Science students only. A STRONG BACKGROUND IN PROBABILITY AND STATISTICS, LINEAR ALGEBRA, AND PYTHON PROGRAMMMING IS ASSUMED. IN ADDITION, BACKGROUND IN MACHINE LEARNING (COMPSCI 589 AND COMPSCI 689) AND NEURAL NETWORKS (COMPSCI 682) IS ASSUMED. SEATS SAVED FOR INCOMING GRADUATE STUDENT REGISTRATION.STUDENTS NEEDING SPECIAL PERMISSION MUST REQUEST OVERRIDES VIA THE ON-LINE FORM: https://www.cics.umass.edu/overrides.