Reverse diffusion from an OU process yields two new MCMC classes (pseudo-marginal and path-space) for sampling unnormalized densities without discretization or score estimation.
Title resolution pending
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
method 1
citation-polarity summary
years
2026 4verdicts
UNVERDICTED 4roles
method 1polarities
use method 1representative citing papers
Introduces global and local contraction coefficients under E_γ-divergence to derive explicit mixing-time bounds for projected Langevin Monte Carlo and independent Metropolis-Hastings, including heavy-tailed cases.
ZOPPA at fixed positive temperature converges under minimal assumptions by acting as an exact proximal point method on a smoothed objective, with explicit connections back to the original function and convergence for its sampled version.
Coupled initial noises in diffusion models, with designed dependence but unchanged marginal Gaussians, improve generated image diversity on Stable Diffusion variants while preserving quality and alignment.
citing papers explorer
-
A New Perspective on Reverse Diffusion for Monte Carlo Sampling
Reverse diffusion from an OU process yields two new MCMC classes (pseudo-marginal and path-space) for sampling unnormalized densities without discretization or score estimation.
-
Local and Global Contraction Principles for MCMC Mixing
Introduces global and local contraction coefficients under E_γ-divergence to derive explicit mixing-time bounds for projected Langevin Monte Carlo and independent Metropolis-Hastings, including heavy-tailed cases.
-
Convergence of zeroth-order proximal point algorithms in the high-temperature regime
ZOPPA at fixed positive temperature converges under minimal assumptions by acting as an exact proximal point method on a smoothed objective, with explicit connections back to the original function and convergence for its sampled version.
-
Couple to Control: Joint Initial Noise Design in Diffusion Models
Coupled initial noises in diffusion models, with designed dependence but unchanged marginal Gaussians, improve generated image diversity on Stable Diffusion variants while preserving quality and alignment.