<ul><li>Diffusion models (DMs) are widely used for representing complex priors in Bayesian inverse problems (BIPs).</li><li>A new ensemble-based algorithm has been proposed to perform posterior sampling for BIPs without using heuristic approximations.</li><li>The algorithm combines DM-based methods with the sequential Monte Carlo (SMC) method.</li><li>Theoretical analysis shows that the proposed algorithm provides more accurate reconstructions in inverse problems in imaging compared to existing DM-based methods.</li></ul>

Solving Inverse Problems via Diffusion-Based Priors: An Approximation-Free Ensemble Sampling Approach

Discover more