Adaptive Importance Sampling for Joint Parameter and Covariance Bayesian Inversion
A Monte Carlo inference scheme for nonlinear multioutput models that jointly estimates model parameters and the noise covariance matrix by splitting the unknowns into two blocks and alternating adaptive importance sampling over parameters with a maximum-likelihood update of the covariance. A recycling step then re-weights the accumulated samples to recover a full Bayesian posterior over the covariance matrix without re-evaluating the costly forward model.
D
Data
Text
Adaptive posterior distributions for uncertainty analysis of covariance matrices in Bayesian
This paper introduces adaptive target adaptive importance sampling (ATAIS), a Monte Carlo scheme for Bayesian inversion problems where one must jointly infer the parameters of a nonlinear multioutput…