Courses Prof. René Vidal has taught at the University of Pennsylvania and Johns Hopkins — spanning generative models, mathematics of deep learning, computer vision, control, and signal processing.
Active courses at Penn.
A graduate course covering the modern landscape of generative modeling — VAEs, GANs, normalizing flows, score-based diffusion models, and their connections to information theory, optimal transport, and dynamical systems.
Prerequisites: probability, linear algebra, and a prior course in machine learning.
Every course taught by Prof. Vidal — newest first.
Three threads that run through the courses.
From overparameterised loss landscapes to neural collapse — a coarse-grained tour of why deep networks generalize, and when they don't.
The mathematical foundations behind modern generative AI — diffusion, flows, score matching, and their training dynamics.
Foundational courses in computer vision, signals & systems, and linear systems theory — the building blocks for downstream research.
Penn ESE/CIS graduate students can register for Deep Generative Models in the Fall.