Closed mkretsch327 closed 3 years ago
Thank you for taking the time to submit a PR!
You are right that in this case the decorator is overkill, yet checking that jax.scipy
implements the right behavior in the CI is also important; so thanks for keeping the test :)
Thank you for adding a new constraint. Besides checking the support, they'll allow to do automatic transformations of the logpdf and automatically choose the most appropriate sampler.
This looks good to me, merging. Thank you !
I noticed unexpected/incorrect behavior of the Poisson distribution- mainly that
dist.Poisson(1).logpdf(0.)
evaluated to-jnp.inf
, rather than -1.This PR updates the support constraint that was causing this at this edge case. Behavior from the
logpmf
function of the poisson distribution inscipy.stats
covers the expected behavior of thelogpmf
function outside of the domain of bothx
andlmbda
.I've added a couple of new unit tests.