Checklist for first release

[ablaom]

I'm keen to register this package and add the models to the MLJ registry so they become discoverable to MLJ users. I know there is work afoot to add documentation, but I think the doc-strings suffice for now.

@EssamWisam Can you please add a checklist for any other issues you feel need addressing before release?

[EssamWisam]

Sorry for the delay. Among other things, have been snowed under with college work which is expected to end soon. Here is the high priority checklist:

• Integrate Entity Embeddings
• Ensure dicts do not force abstract types #20
• Fix union type hack #6
• When creating new columns use category names instead of numbers #10

I am close to being done for the first item (integrating entity embeddings).

The broader checklist, includes the other issues and adding the following tutorials to the documentation (as well as sprucing it up as needed):

• A tutorial for entity embeddings (e.g., with some plots to showcase what it does)
• A tutorial on standardization and how it can make a huge difference in performance
• A tutorial comparing classical/widely known encoders together
• Potentially, a tutorial to analyze overfitting in target encoding
• A tutorial to compare different contrast encoders (if they do show different performance) else just using them
• A tutorial comparing one-hot encoder only with one-hot encoder preceded by a cardinality reducer (to see an improvement in runtime and potentially performance)

[ablaom]

Hi @EssamWisam. Any recent progress here?

[EssamWisam]

Hi @EssamWisam. Any recent progress here?

Finished the most consequential step of integrating entity embeddings. Expect more progress soon.

[ablaom]

@EssamWisam Can we please prioritize registration of this package and leave all those tutorials for later? This great functionality is still not available to our users.

[EssamWisam]

I think I need I just need #5 now and finishing up docs if that is also needed for registration (maybe not as MLJ docstrings will propagate). I aim to finish at least one of these in the next two days.

[ablaom]

| docs if that is also needed for registration

Not needed for registration, but do merge what you have so far.

And, yes, the model docstrings should propagate after the next MLJ Model Registry update, which requires run update and tag a new release at MLJModels, which I can do. Perhaps add this step to the checklist, or open a new issue, so I don't forget.

[EssamWisam]

@ablaom after merging #27 you can:

• Go ahead and merge
• Update compat bounds
• Register package (pending)
• Ensure the documentations propagates to MLJ by updating MLJModels

[EssamWisam]

In about a week I will have finished the initial comprehensive documentation (minus the tutorials) and then will add the tutorials on my own pace inshallah.

[ablaom]

To clarify, you are happy for me to register without merging #19? I thought perhaps we could address the minor issues raised in review there and merge as is, no?

[EssamWisam]

I am indifferent towards registering without merging #19 but I do expect I finalize a version suitable for users by next Monday. So if you feel the necessity of exposing the methods in MLJ Docs before that happens then do publish before we merge it.

[ablaom]

Okay, I'm going ahead and registering now.

[ablaom]

Hang on, I can see there are some compat updates to deal with first. Ammending checklist above.

[EssamWisam]

Thank you. I couldn't get to this the last time I scheduled working on it, but will boost it's priority to finish before next Monday.