Generalization in Diffusion Models Arises from Geometry-Adaptive Harmonic Representations

Generalization in Diffusion Models Arises from Geometry-Adaptive Harmonic Representations

Explore a comprehensive lecture on the topic of generalization in diffusion models and their connection to geometry-adaptive harmonic representations. Delivered by Zahra Kadkhodaie from New York University, this one-hour and four-minute talk delves into the emerging generalization settings within the field. Gain insights into how diffusion models achieve generalization through their unique ability to adapt to geometric structures, and understand the implications of these findings for the broader field of machine learning and artificial intelligence.