Noise, Denoise, Correct: MCMC Posterior Sampling with Diffusion Priors in Three Steps

Name: Noise, Denoise, Correct: MCMC Posterior Sampling with Diffusion Priors in Three Steps
Uploaded: 2026-10-18T05:01:16.730769+00:00
Duration: 353 s
Description: Diffusion waltz is an MCMC method for exact posterior sampling using SDEdit-style noising-denoising proposals corrected by Metropolis–Hastings. It handles non-differentiable models, like Navier–Stokes initial condition recovery, and incorporates observation information via conditional noising to improve mixing and performance compared to standard methods.

Diffusion waltz is an MCMC method for exact posterior sampling using SDEdit-style noising-denoising proposals corrected by Metropolis–Hastings. It handles non-differentiable models, like Navier–Stokes initial condition recovery, and incorporates observation information via conditional noising to improve mixing and performance compared to standard methods.

Open the full episode