Open kMutagene opened 2 years ago
After some discussion, we arrived at the use of 2 functions for this problem, a performant version that does not do these checks (prefixed by _
), and a checked version (e.g. _gamma
and gamma
). the unchecked versions should be used internally.
I am not sure if this can be labeled as a bug, please adjust accordingly. Functions that are approximated (such as gamma, erfcx, etc.) may return incorrect results for edge cases such as +/- infinity.
Example: gamma(infinity) should be infinity, but returns nan via approximation:
https://www.wolframalpha.com/input?i=gamma%28infinity%29
My suggestion would be catching edge cases on the input, so in the case of the gamma approximation: