1: \begin{abstract}. We study the problem of sampling high and infinite dimensional target

2: measures arising in applications such as conditioned diffusions and

3: inverse problems. We focus on those that arise from approximating

4: measures on Hilbert spaces defined via a density with respect to a

5: Gaussian reference measure. We consider the Metropolis-Hastings algorithm

6: that adds an accept-reject mechanism to a Markov chain proposal in

7: order to make the chain reversible with respect to the the target measure. 

8: We focus on cases where the proposal is either a Gaussian random walk

9: (RWM) with covariance equal to that of the reference measure or an

10: Ornstein-Uhlenbeck proposal (pCN) for which the reference measure

11: is invariant. 

12: 

13: Previous results in terms of scaling and diffusion limits suggested

14: that the pCN has a convergence rate that is independent of the dimension

15: while the RWM method has undesirable dimension-dependent behaviour.

16: We confirm this claim by exhibiting a dimension-indepen\-dent Wasserstein

17: spectral gap for pCN algorithm for a large class of target measures.

18: In our setting this Wasserstein spectral gap implies an $L^{2}$-spectral

19: gap. We use both spectral gaps to show that the ergodic average satisfies

20: a strong law of large numbers, the central limit theorem and non-asymptotic

21: bounds on the mean square error, all dimension independent. In contrast

22: we show that the spectral gap of the

23: RWM algorithm applied to the reference measures degenerates

24: as the dimension tends to infinity. 

25: \end{abstract}

26: