Id: arxiv/2607.29525 Url: https://arxiv.org/abs/2607.29525 Tags: math.OC, cs.IT Score: 0 Title: Physics Matters in PnP: Recovery Guarantees with the MMSE and NN Denoisers Domain: math.OC Source: arxiv Authors: Tobias Wolf, Jalal Fadili, Jin Guo, Roy Y. He Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-07-31 Page: https://delphi.science/r/arxiv/2607.29525 We investigate the forward-backward-splitting version of the Plug and Play (PnP) method for linear ill-posed problems with MMSE estimators as denoisers. In contrast to existing literature, we consider estimators which are specialized for (degenerate) Gaussian noise with possibly non-diagonal covariance matrices. We further deviate from the classical iteration by replacing parts of the descent step with a linear operator that relates the observation noise to that of the MMSE estimator. Under mild assumptions, we derive several properties of the denoiser and prove recovery guarantees of the…