Fix fast_random_hypergraph

[nwlandry]

fast_random_hypergraph suffers from the same errors that were pointed out in issue #652. This attempts to fix these errors. As a side note, this now converts sampling using the log method into np.random.geometric(). Not sure how it compares performance-wise, but would be interesting to see.

[nwlandry]

The only thing that doesn't work is the seed argument for numpy.random.geometric. Any ideas @leotrs @tlarock @maximelucas?

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.50%. Comparing base (69ff425) to head (e5bd37f).
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #655      +/-   ##
==========================================
+ Coverage   93.44%   93.50%   +0.06%     
==========================================
  Files          64       64              
  Lines        5004     5005       +1     
==========================================
+ Hits         4676     4680       +4     
+ Misses        328      325       -3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[maximelucas]

Thanks! Maybe add a test to check that (i) the seed works now (the reproducibility part) and (ii) fast_random_hypergraph doesn't crash with seed=5 anymore.

[nwlandry]

@maximelucas, I added documentation to the new method I made and I added a test. (I verified that the test fails on the main branch)

[nwlandry]

Quick aside is that comparing np.random.geometric vs. xgi.geometric, the new xgi version version is 4x faster.

[nwlandry]

Okay, I fixed a bunch of random errors and should be good to go @maximelucas!

[maximelucas]

Quick aside is that comparing np.random.geometric vs. xgi.geometric, the new xgi version version is 4x faster.

Yea often numpy functions are much faster when generating a bunch of random numbers but slower for just one.

[tlarock]

Quick aside is that comparing np.random.geometric vs. xgi.geometric, the new xgi version version is 4x faster.

Faster is great! Do they give the same distribution?

[nwlandry]

I can do a KS test at some point!

[nwlandry]

Okay, the distributions look similar, and the methods are the same (almost), but I think the difference that the KS test picked up on is due to numerical errors in the evaluation of the log function.