MA4b.3

The Geometry of Noise: Why Diffusion Models Don't Need Noise Conditioning

Mojtaba Sahraee-Ardakan, Mauricio Delbracio, Peyman Milanfar, Google, United States
Session:
MA4b: Generative Modeling for Inverse Problems Lecture
Track:
Adaptive Systems, Machine Learning, and Data Analytics
Location:
Room 4
Presentation Time:
Mon, 26 Oct, 11:05 - 11:30 PT (UTC -7)
Session Co-Chairs:
Yu Sun, Johns Hopkins University, United States and Qing Qu, University of Michigan, United States
Presentation
Discussion
Resources
No resources available.
Session MA4b
MA4b.1: Global Attention with Linear Complexity for Exascale Generative Data Assimilation in Earth System Prediction
Guannan Zhang, Oak Ridge National Laboratory, United States
MA4b.2: Title: Beyond scores: proximal diffusion models
Zhenghan Fang, Mateo Diaz, Johns Hopkins University, United States; Sam Buchanan, UC Berkeley, United States; Jeremias Sulam, Johns Hopkins University, United States
MA4b.3: The Geometry of Noise: Why Diffusion Models Don't Need Noise Conditioning
Mojtaba Sahraee-Ardakan, Mauricio Delbracio, Peyman Milanfar, Google, United States
MA4b.4: Conservation-constrained diffusion maps for dynamic radiography in hydrodynamics
Aviral Prakash, Khoa Nguyen, Daniel Serino, Marc Klasky, Los Alamos National Laboratory, United States
Contacts