Fast Bayesian inversion for high dimensional inverse problems

We investigate the use of learning approaches to handle Bayesian inverse problems in a computationally efficient way when signals be inverted present moderately high number dimensions and are large number. propose tractable regression approach which has advantage produce full probability distributions as approximations target posterior distributions. In addition provide confidence indices on predictions, these allow better exploration multiple equivalent solutions exist. then show how can used for further refined predictions using importance sampling, while also providing carry out uncertainty level estimation if necessary. The relevance proposed is illustrated both simulated real data context physical model inversion planetary remote sensing. shows interesting capabilities terms computational efficiency multimodal inference.