Closed AlexAndorra closed 6 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 91.84%. Comparing base (
244fb97
) to head (2c279ba
).
Failing test seems completely unrelated. Should I just rerun it?
Empirical on my side, but I know Bill also told me that. I'm guessing this is similar to the fact that the centered Normal parametrization works better in a hierarchical model for groups which have a lot of data
El El lun, 11 mar 2024 a la(s) 17:09, Juan Orduz @.***> escribió:
@.**** commented on this pull request.
In pymc/gp/hsgp_approx.py https://github.com/pymc-devs/pymc/pull/7189#discussion_r1520368416:
The centered approximation can be more efficient when
the GP is stronger than the noise
beta = pm.Normal("beta", sigma=sqrt_psd, size=gp._m_star)
f = pm.Deterministic("f", phi @ beta)
Out of curiosity: is this something empirical, or is there a general statement about it? :)
— Reply to this email directly, view it on GitHub https://github.com/pymc-devs/pymc/pull/7189#pullrequestreview-1929102936, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHIJMTD5QI4GU7JWS4D4X7TYXYFNZAVCNFSM6AAAAABEPONYL6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMYTSMRZGEYDEOJTGY . You are receiving this because you were assigned.Message ID: @.***>
@AlexAndorra small nitpick, would have been better to have two separate commits for the two unrelated changes (Mixture, HSGP)
True. Noted for next time @ricardoV94
Description
Just a small PR to improve and fix some typos in the doc pages of:
prior_linearized
method)ZeroInflated
distributions (switched from "variates" to "draws", which is more common and much clearer when teaching)Hurdle
distributions (same change as previous point)Ready for review and merge
Checklist
Type of change
📚 Documentation preview 📚: https://pymc--7189.org.readthedocs.build/en/7189/