How to add icdf to a truncated normal?

numpyro
1

Hello friends, I hope you are well. I would like to use the wrapper for jaxns with a prior that uses a truncated normal, but I’m running into this error:

    519 def icdf(self, q):
    520     """
    521     The inverse cumulative distribution function of this distribution.
    522 
    523     :param q: quantile values, should belong to [0, 1].
    524     :return: the samples whose cdf values equals to `q`.
    525     """
--> 526     raise NotImplementedError

NotImplementedError:

Which states that the truncated normal has not implemented icdf method. What would be the best way to go around this problem? Is there maybe a way of building a truncated normal using transformations that would take advantage of the normal icdf? thanks a lot!

2

Nevermind, using a uniform prior solved the issue (blush)

offst = numpyro.sample('offst', dist.Uniform(-1, 1))