DOC Feature highlights for 0.23 #17062
Conversation
Co-Authored-By: Christian Lorentzen <lorentzen.ch@googlemail.com>
Co-Authored-By: Christian Lorentzen <lorentzen.ch@googlemail.com>
|
I think the positional deprecation also deserves to be here. |
I'd be fine with this focusing on enhancements and new features. HTML vis belongs here when merged. sample_weight in Lasso and ElasticNet? Changing away from Boston in examples? (a good media piece?) Otherwise I suspect we're not going to list much more for this release. A lot of work behind the scenes for this one!! |
Co-Authored-By: Joel Nothman <joel.nothman@gmail.com>
…lease_highlights_023
…-learn into release_highlights_023
| print(reg.score(X_test, y_test)) | ||
|
|
||
| ############################################################################## | ||
| # Finally, users can now define :ref:`monotonic constraints |
There was a problem hiding this comment.
Just for transparency, monotonic cst are in an example of their own after Joel's suggestion. It's not me trying to push my own PRs
|
Thanks Joel, I added a section about Lasso and ENet and separated the mnt_cst Re positional stuff: I agree with Joel that this isn't really a feature. I feel like the top entry in the what's new would be enough Re usage of Boston dataset: same, I feel like this would be a nice top-entry in the what's new, but not sure about including it in the highlights? |
|
@NicolasHug, apparently people prefer the |
|
yes, same for #16648 I'll update when / if the PR gets merge. I personally prefer the current form |
|
I'd also happy to have the boston dataset change as a highlight here. |
|
The kwonly story is one of "Helping our users write readable code". The
Boston story is one of "Ending a racist legacy in the teaching of data
analysis"... But both these stories may be better told in a blog, @cmarmo?
|
Stories are always better told in a blog than in a software documentation, even if it is the best documentation in the world like yours ... ;) |
My understanding was that if the foundation website is to have an article similar to this one for 0.23, then it should make clear that the "official" highlights are in the gallery, and link to it I personally agree a blog post is better suited both for Boston and for the kwonly. |
|
This needs a section on diagram display.
|
|
Comments addressed and added entry for html repr |
|
|
||
| clf = make_pipeline(preprocessor, LogisticRegression()) | ||
| clf | ||
|
|
There was a problem hiding this comment.
Might be nice to demonstrate a composite gaussian processes kernel too??
There was a problem hiding this comment.
maybe this can be done in the docs instead? I feel like a single example is enough, and we want the highlights to be short
Co-authored-by: Joel Nothman <joel.nothman@gmail.com>
|
@ogrisel @thomasjpfan @rth @adrinjalali @glemaitre @jeremiedbb @TomDLT I think we can get this in now |
…lease_highlights_023
…-learn into release_highlights_023
|
Looks good to me. Tiny nitpicks: I wouldn't do plt.legend(); plt.show() on the same line. It's not advised in Python styling. |
Co-authored-by: Gael Varoquaux <gael.varoquaux@normalesup.org>
|
@adrinjalali I'm not sure what PR I should ping anymore but this should be in the release ;) |
|
I guess we can continue pinging #17010 |
CC @scikit-learn/core-devs please review or directly edit and add new entries as you please
CC @cmarmo too