BriefGPT.xyz
Mar, 2024
分治后验采样用于去噪扩散先验
Divide-and-Conquer Posterior Sampling for Denoising Diffusion Priors
HTML
PDF
Yazid Janati, Alain Durmus, Eric Moulines, Jimmy Olsson
TL;DR
利用去噪扩散模型（DDM）作为先验来解决逆贝叶斯问题的兴趣最近显著增加。本研究采用不同方法，利用DDM先验的特定结构定义了一组中间和简化的后验采样问题，相比以前的方法，降低了近似误差。我们通过使用合成示例和各种图像恢复任务来经验性地展示了我们方法的重建能力。
Abstract
Interest in the use of
denoising diffusion models
(DDM) as priors for solving
inverse bayesian problems
has recently increased significantly. However, sampling from the resulting
→