GP-Latent update for v4

[bwengals]

Associated with #72, and is tied to PR in pymc-devs/pymc #5055.

The notebook was reran using the fixes from the v4 GP branch, so the examples will error if merged before pymc v4 is done. Also I tried to make the right changes to be aligned with the style guide, but please let me know if I missed anything!

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2021-10-19T06:52:22Z
----------------------------------------------------------------

the tags are not working, I think it's because of the raw cell, try putting it right below the title, in the markdown cell as this syntax is markdown for sphinx

fonnesbeck commented on 2022-04-02T19:52:36Z
----------------------------------------------------------------

@bwengals are you able to fix this?

[fonnesbeck]

@bwengals are you able to fix this?

---

View entire conversation on ReviewNB

[fonnesbeck]

Please re-run the notebook to get rid of the errors in the current render.

[fonnesbeck]

The fit is not great on the last model. Maybe boost the tune interval to 2000?

[fonnesbeck]

Are you using a bespoke invlogit instead of pm.math.invlogit because of stability issues?

[ricardoV94]

Are you using a bespoke invlogit instead of pm.math.invlogit because of stability issues?

If that's the case, check if it is still needed. We have improved invlogit stability in Aesara. Also the Bernoulli can take logit_p directly

[fonnesbeck]

@bwengals how close is this one to being ready?

[bwengals]

So sorry for missing these comments :( I did finish it up today, hopefully I addressed you guys' comments. Unfortunately I used the wrong email address for a few previous commits, then made a big git mess trying to change them. If you guys don't mind, I'm just going to open a new PR tomorrow...

[bwengals]

Closing for #371