arXiv:2510.12691v2 Announce Type: replace-cross
Abstract: Diffusion models have emerged as powerful generative priors for high-dimensional inverse problems, yet learning them when only corrupted or noisy observations are available remains challenging. In this work, we propose a new method for training diffusion models with Expectation-Maximization (EM) from corrupted data. Our proposed method, DiffEM, utilizes conditional diffusion models to reconstruct clean data from observations in the E-step, and then uses the reconstructed data to refine the conditional diffusion model in the M-step. Theoretically, we provide monotonic convergence guarantees for the DiffEM iteration, assuming appropriate statistical conditions. We demonstrate the effectiveness of our approach through experiments on various image reconstruction tasks.
Magnetoencephalography reveals adaptive neural reorganization maintaining lexical-semantic proficiency in healthy aging
Although semantic cognition remains behaviorally stable with age, neuroimaging studies report age-related alterations in response to semantic context. We aimed to reconcile these inconsistent findings