Add logistic regression example with NUTS

[rlouf]

This PR adds an example of a logistic regression model, similar to PyMC3's, and samples from its posterior with NUTS.

[codecov]

Codecov Report

Merging #23 (3fd6aea) into main (d321748) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main       #23   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            9         9           
  Lines          344       345    +1     
  Branches        14        14           
=========================================
+ Hits           344       345    +1     

Impacted Files	Coverage Δ
aehmc/nuts.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d321748...3fd6aea. Read the comment docs.

[twiecki]

@rlouf Anything that's missing here?

[rlouf]

Yep, correct results.

[brandonwillard]

What's causing the current issue with the logistic regression results?

[rlouf]

The mass matrix + step size that is used by PyMC3 leads to only divergent transitions; actually the largest step size at which we get an acceptance rate > .5 is orders of magnitude smaller than that used by PyMC3. Investigating why that's the case.

[twiecki]

@rlouf Which mass matrix initialization method is this using? PyMC3's or Stan's? If PyMC3, @ricardoV94 actually found a problem there and @aseyboldt has worked on a new one that works much better.

[twiecki]

pymc-devs/pymc#5004

[rlouf]

Ah thanks, I'll check this out. I decided to check against numpyro's returned parameters in the end.

Nevertheless since PyMC3 samples fine with the parameters it obtains from warmup I shouldn't have any issue on my end.

[twiecki]

@rlouf Can we add reviewing here as well?

[rlouf]

Do you mean NBReview by reviewing?

[twiecki]

Yeah.

…

On Mon, Sep 20, 2021 at 2:53 PM Rémi Louf ***@***.***> wrote: Do you mean NBReview by reviewing? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#23 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFETGHCM4BZUMDBQ7FVU43UC44D7ANCNFSM5DJJB4SA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[rlouf]

I guess that's @brandonwillard's decision to make.

[brandonwillard]

Do we need to add an application to the aesara-devs org or something?

[rlouf]

Starting over. Closing this PR and opening a new one.