search

acerbilab / pyvbmc

PyVBMC: Variational Bayesian Monte Carlo algorithm for posterior and model inference in Python
https://acerbilab.github.io/pyvbmc/
BSD 3-Clause "New" or "Revised" License
114 stars 6 forks source link

Bugs on `plb` and `pub` #97

Closed pipme closed 2 years ago

pipme commented 2 years ago

plb and pub are in tranformed space, i.e. unconstrained space. plb_orig and pub_orig are in original constrained space.

https://github.com/lacerbi/pyvbmc/blob/4a071d8c44e38bade6eb0709cfbed3a3b7880801/pyvbmc/vbmc/gaussian_process_train.py#L417-L429

The usage of plb and pub above are wrong. We don't need to parse plb and pub to train_gp function actually. Check other parts that have similar bugs. I will add a fix on sparse-develop branch first.

lacerbi commented 2 years ago

As a further comment, the issue above arises because train_gp takes as input self.plausible_lower_bounds and self.plausible_upper_bounds which are in original space (as far as we can tell), while here the passed bounds should be in internal space.