python 3.x - Can tensorflow probability bijectors be used to supply parameters to the base distribution?

Question

Usually when working with normalising flows, the base distribution is a standard normal distribution. However, I'm currently trying to use tensorflow probability to construct a bijector working with a student t distribution as a base distribution.

Is it possible to include the estimation of the degrees of freedom of that base distribution in the bijector itself? Or does the implementation in tensorflow probability not allow for base distributions to not be fully parameterised?

question from:https://stackoverflow.com/questions/66066426/can-tensorflow-probability-bijectors-be-used-to-supply-parameters-to-the-base-di

Answer 1

Waitting for answers