Seminars and Events
Diffusion, Probability, and the Age of Creative AI
Denoising diffusion models have spurred significant gains in density modeling and image generation, precipitating an industrial revolution in text-guided AI art generation. I’ll give a high level overview of developments in the field before discussing a new information-theoretic foundation for diffusion models which provides an exact equivalence between probability and optimal denoising. This insight leads to several improvements over traditional diffusion models including improved density estimation, interpretability through information decomposition, and faster sampling.
Greg Ver Steeg recently started as an associate professor in CS at the University of California Riverside, after many wonderful years working at ISI. His research incorporates ideas from physics and information theory in machine learning in order to understand complex systems like human behavior and biology.
YOU ONLY NEED TO REGISTER ONCE TO ATTEND THE ENTIRE SERIES – We will send you email announcements with details of the upcoming speakers.
Register in advance for this webinar: https://usc.zoom.us/webinar/register/WN__0VhakI6Q6i3JsasdmNWcA
After registering, you will receive an email confirmation containing information about joining the Zoom webinar.
If speaker approves to be recorded for this AI Seminar talk, it will be posted on our USC/ISI YouTube page within 1-2 business days: https://www.youtube.com/user/USCISI.
Host: Jay Pujara, POC: Pete Zamar