New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

RNG statistical correctness checks for floating point and Gaussian values #85

Open
kyonifer opened this Issue Oct 30, 2018 · 2 comments

Comments

Projects
None yet
2 participants
@kyonifer
Owner

kyonifer commented Oct 30, 2018

It would be useful to implement some tests to check the rigor of our Gaussian and floating point random number streams, since the underlying PCG implementation we are based on only generates integral types (and thus our 1:1 comparisons to them are only for the integral types, not our derived representations). We could potentially compare ourselves to another jvm implementation of PCG or call into a test suite.

@peastman

This comment has been minimized.

Contributor

peastman commented Oct 31, 2018

testFloatDistribution() and testGaussianDistribution() do check whether the distributions are correct. They aren't super robust, but if there were a bug in the implementation of Box-Muller it would be very unlikely to pass.

@kyonifer

This comment has been minimized.

Owner

kyonifer commented Oct 31, 2018

Right... this is referring to the potential need for more rigorous testing. We already have that for integers via the direct comparison to the reference implementation which was itself tested against TestU01. But we don't have that for other (derived from integral) types because the reference implementation doesn't generate them.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment