Bayesian Inverse Problems
Overview
Bayesian inverse problems use prior knowledge and measurement models to recover hidden images, fields, or physical parameters, while representing ambiguity with a posterior distribution. Common approaches include MCMC and variational inference, simulation-based inference, surrogate forward models, and learned priors based on diffusion models, normalizing flows, or neural operators. Applications range from MRI and image restoration to seismic tomography and PDE-constrained inference, where computation can be costly and the posterior may be high-dimensional or multimodal. A central challenge is making inference faster without sacrificing calibrated uncertainty: good measurement fit or reconstruction scores alone do not ensure that posterior samples capture the right spread and plausible alternatives.
Latest papers
128 papers14 papers in the last four weeks, down 22% on the four weeks before. 0.1% of all new papers.
ShuffleFlow: Scalable Posterior Inference for Bayesian Inverse Imaging
Neural network surrogates with uncertainty quantification for inverse problems in partial differential equations
Trustworthy MRI Reconstruction via Bayesian Uncertainty Quantification with Sparsity Prior Models
Amortized mean-shift interacting particles
Probabilistic Signature Inversion: Learning Conditional Distributions from Truncated Signatures
Bridging data-driven priors via the score function for posterior sampling -- Comparative review and experimental study
A Stabilized Path-Space Approach to Diffusion-Based Posterior Sampling
Learning the Universe: Posterior Reliability of Neural Generative Models in High-Dimensional Field-Level Inference of Cosmic Initial Conditions
Bridged SBI: Correcting Biased Low-Fidelity Posteriors for Cost-Efficient High-Fidelity Inference
Structure-Preserving Correction Learning for Sparse Bayesian Inference in Brain Source Imaging
Generalized TV-- Structured Priors for Bayesian Mapping
The Right Measure for Physics-Constrained Generation: A Co-Area Correction for Posterior-Consistent PDE Inverse Problems
satisfying the physics'' is not the same as sampling the posterior,'' and give a principled correction for uncertainty-aware scientific inference.