Open lrnv opened 11 months ago
There are analytical formulas for the derivatives. One of them is fine, but the other involves the Meijer G function, which we don't have (#265).
We could add a chain rule for the derivative with respect to the second argument, but ForwardDiff doesn't support ChainRules (unless you also import https://github.com/ThummeTo/ForwardDiffChainRules.jl, which we don't currently do here).
It looks like the derivatives you linked to on wikipedia are for the gamma incomplete function, and not its inverse, if I am not mistaking.
I am hitting the following issue :
Is there somewhere an implementation for such a derivative ? I guess the routines for the inverse of the gamma incomplete function are based on numerical root finding, so it that even possible ?