The problem of Metropolis is that in high dimension the random walk is nearly impossible to explore the posterior space properly in finite time. For a complex system like PDE, I would suggest you to study the posts by @aseyboldt (see more details here Using pymc3 NUTS sampler for an "external model" having its derivatives with respect to its parameters). In many cases, gradient could be approximated which allow you to use NUTS.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| NUTS Sampler: Effective samples is smaller than 200 for some parameters | 4 | 5073 | July 6, 2020 | |
| Any possible ways to get faster ways of pm.sample() | 6 | 1812 | October 20, 2022 | |
| The number of effective samples is smaller than 25% for some parameters | 4 | 11458 | April 5, 2018 | |
| PyMC3 slows rapidly with increasing numbers of parameters | 5 | 11197 | December 3, 2018 | |
| Metropolis: ideal balance between slow-mixing and high rejection rate? | 9 | 1930 | May 16, 2018 |