Skip to content
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

Update RNG usage with numpy's new interface #108

Draft
wants to merge 7 commits into
base: master
Choose a base branch
from

Conversation

clbarnes
Copy link
Contributor

@clbarnes clbarnes commented Sep 8, 2022

get_rng is now just a thin wrapper around numpy.random.default_rng, which reduces the set of types which could be used as a seed but improves consistency. default_rng may change the underlying algorithm in future numpy releases, which means that results are not necessarily stable for the same seed across numpy versions; as such, users can pass their BitGenerator of choice in the seed argument, and tests are hard-coded to use today's preferred algorithm.

Updating from the deprecated RNG (where the algorithm wasn't even selectable) to today's preferred one required changing all of the hardcoded reference values in the tests. For most, this was a pretty minor difference within reason for changing the random seed, but in a couple of cases the change was more significant. I'm not sure how best to check whether this is "real" or not.

Additionally removed EOL python versions from tox and added more recent releases.

Travis may be dead so I'll look into github actions instead.

@clbarnes
Copy link
Contributor Author

clbarnes commented Sep 8, 2022

Rebased on the GHA branch used in #109

numpy.random.default_rng may use a different algorithm between versions,
so testing with a specific algorithm is best for stability.
This necessitates updating all of the hard-coded test values.
Most are basically the same, but a few are not.
assert np.allclose(q, .32697921)
_, q = bct.modularity_louvain_dir(x, seed=stable_rng)
# assert np.allclose(q, .32697921) # legacy np.random.RandomState
assert np.allclose(q, 0.373_475_890) # stable_rng
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Result with new RNG is quite different to previous.

assert np.allclose(q, .07885327)
# N.B. this result is quite different
# assert np.allclose(q, .07885327) # legacy numpy.random.RandomState
assert np.allclose(q, 0.114_732_792) # stable_rng
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Result with new RNG is quite different to previous.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant