1: \begin{abstract}

2: 

3: 

4: Recent advancements in Markov chain Monte Carlo (MCMC) sampling and surrogate modelling have significantly enhanced the feasibility of Bayesian analysis across engineering fields.

5: However, the selection and integration of surrogate models and cutting-edge MCMC algorithms, often depend on ad-hoc decisions.

6: A systematic assessment of their combined influence on analytical accuracy and efficiency is notably lacking.

7: The present work offers a comprehensive comparative study, employing a scalable case study in computational mechanics focused on the inference of spatially varying material parameters, that sheds light on the impact of methodological choices for surrogate modelling and sampling.

8: We show that a priori training of the surrogate model introduces large errors in the posterior estimation even in low to moderate dimensions.

9: We introduce a simple active learning strategy based on the path of the MCMC algorithm that is superior to all a priori trained models, and determine its training data requirements.

10: We demonstrate that the choice of the MCMC algorithm has only a small influence on the amount of training data but no significant influence on the accuracy of the resulting surrogate model.

11: Further, we show that the accuracy of the posterior estimation largely depends on the surrogate model, but not even a tailored surrogate guarantees convergence of the MCMC.

12: Finally, we identify the forward model as the bottleneck in the inference process, not the MCMC algorithm.

13: While related works focus on employing advanced MCMC algorithms, we demonstrate that the training data requirements render the surrogate modelling approach infeasible before the benefits of these gradient-based MCMC algorithms on cheap models can be reaped.

14: 

15: \end{abstract}

16: