Closed dstansby closed 1 year ago
Had a think about this for a bit, and for now I think it's easiest to leave this on the back burner as a potential future development. For ease we should just use a non-differentiable MCMC implementation to get started.
Had more of a think about this, and it's definitely differentiable 👍
Some MCMC packages make use of posterior probability functions that can be differentiated to speed up evaluation. Need to understand if the posterior probability function for DEM inversions is differentiable.