Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a direct intermediate initialization strategy for Gaussian-tilted diffusion posterior samplers, specifically MCGDiff, by analytically transporting samples from a softened clean-space posterior via a Gaussian bridge to the noisy-space target. This approach leverages the structure of the pulled-back problem, where effective observation variance can be significantly higher than diffusion noise, to improve finite-particle performance at the cost of asymptotic consistency. The hybrid method, using moment-matching posterior sampling for the initial solver, achieves substantial improvements in sliced Wasserstein distance on inverse problems, particularly when the posterior-relevant mode is rare under the prior.
A novel initialization technique for diffusion samplers bypasses initial steps by directly sampling a softened clean-space posterior, achieving order-of-magnitude gains on rare-mode inverse problems.
Some diffusion posterior samplers construct Gaussian-tilted intermediate distributions along the reverse process. We observe that these targets can be pulled back to clean-space posteriors with weaker conditioning, with samples transported analytically to the corresponding noisy-space target through a Gaussian bridge. For the sequential Monte Carlo (SMC) sampler MCGDiff, the effective observation variance of this pulled-back problem is up to twice the diffusion-noise variance. We exploit this structure to initialize MCGDiff directly at an intermediate time: an approximate solver samples the softened clean-space posterior, the Gaussian bridge maps these samples to the tilted target, and only the remaining SMC suffix is run. This trades asymptotic consistency for finite-particle performance. With moment-matching posterior sampling (MMPS) as the solver, the hybrid improves sliced Wasserstein distance by roughly $2\times$ at matched particle count on a structured Gaussian-mixture inverse problem, and by more than an order of magnitude when the posterior-relevant mode is rare under the prior. A prior-initialization control, which retains the bridge but drops the clean-space conditioning, shows that on MCGDiff's standard Gaussian-mixture benchmark most of the improvement is insensitive to the conditioning. Conditioning the initialization gives a further consistent gain on the structured problem, and becomes decisive on a rare-mode problem, where resampling cannot repopulate a mode absent from the initial population.