AINeutralarXiv – CS AI · 10h ago6/10
🧠
Outlier-Robust Diffusion Solvers for Inverse Problems
Researchers have developed an improved diffusion model-based approach for solving inverse problems that demonstrates robustness to outliers in real-world measurements. The method combines explicit noise estimation, Huber loss optimization, and conjugate gradient methods to outperform existing diffusion model techniques across linear and nonlinear tasks.