On 09/29/2016 11:07 PM, Joel Nothman wrote:
(this has been in drafts a few days and I'm sure there's plenty I've
missed from the lists below)
Well done, everyone! The size of this release - and the group of
people that contributed to it - is even a bit overwhelming. Thanks for
managing the release, Andy... and writing it up as a book!
Thank you for your incredible dedication!
We've got a lot in the works already for 0.19.
There are a number of things that have been a long time coming and
which I'd really like to see happen, such as:
* multiple metrics for cross validation (#7388 et al.)
* documenting and officially making (most) utils public (#6616)
* indicator features for Imputer, done right (#6556)
* KNN imputation (#2989, #4844)
* ColumnTransformer or similar for heterogeneous data (#2034, #886)
* dataset resampling (#1454)
* string handling in OneHotEncoder (#7327)
* interpolation in average_precision_score (#7356)
* tree categorical splits (#4899)
* k-best feature selection from a model's feature_importances_ (#6717)
* ? feature name transformation (#6425)
* ? sample_weight support in CV scoring (#1179, #2879, #3524, #1574;
perhaps this isn't as easy as it looks)
There are things that are important but will probably require more work:
* making common tests and their exceptions more general (perhaps by
way of "estimator tags")
* improving our LSH offerings and integration
It's good to see that you're excited about the same things as me.
I also want to do the numpy-doc update, as it gives SOOO much better
error messages now.
I'll try to put some time into the public utils soon, and I think the
interpolation in average precision is basically done!
I think many of the other things you mentioned are already well on their
way, and maybe we can get 0.19 out within the next 4 month,
to get back on a more regular schedule.
_______________________________________________
scikit-learn mailing list
[email protected]
https://mail.python.org/mailman/listinfo/scikit-learn
