On a generalization of the preconditioned Crank-Nicolson metropolis algorithm (Q1750384)

Approximation of sampling of probability measures on infinite dimensional spaces using Metropolis algorithms on infinite dimensional Hilbert spaces are considered and a generalization of the preconditioned Crank-Nicolson (pCN) proposal is introduced which aim to adopt to the covariance structure of the target measure. The gpCN proposal is motivated and it is shown that it is well defined in function spaces and illustrate its superior performance in a simple but common setting. Moreover, a qualitative convergence result is provided by a comparison argument for spectral gaps.