Setting seed for reproduction and extracting MAP #22
Description
Thank you very much for developing this tool for flexible Bayesian inversion.
I have question about the possibility of setting seed when running sampling through inversion.run().
Up to my experience, it seems that the sampled results differ everytime I run the inversion, which is expected, but for reproducibility, I was looking for an option to set the seed, but could not find it.
Is such an option implemented or is there any way to do so?
Besides, I am also curious if there is way to extract maximum a posteriori (MAP) solution from the sampled models.
Sincerely yours,
Minseong Seo.
Added on 2025/09/08
I found that there is option walkers_starting_states for bayesbay.BaseBayesianInversion and this can initialize the model for each chain which sampling starts from. However, still, it seems that sampling itself is random for every run.