(Maths-ia) Using Score-based generative models as priors for ill-posed inverse problems

Inverse problems are a common class of problems arising in many scientific domains, where well known examples are super-resolution, inpainting, Geosismic imaging, and medical imaging. In this primer, I will review the current litterature on using (score-based) generative models as informative priors to solve such problems and show how this approach was used to solve some problems from Electrocardiogram Analysis (Denoising, missing lead completion and anomaly detection).