Skip to content

Rework fake sampler

Colm Talbot requested to merge update-fake-sampler into master

The fake sampler has been in need of some TLC for a while. This MR cleans up the implementation in a few ways.

  • use_ratio is fully supported now.
  • the plotting of the weights in the example is now in the main code.
  • more details are included in the documentation.
  • changes to the prior will be caught and included in the reweighting.
  • the posterior samples are rejection sampled using the computed weights.
  • the evidence/bayes factor stored in the output result are updated along with an uncertainty estimate.
  • if no likelihood is passed, just reweight based on the prior.

Merge request reports