DOC Release dates for 0.20RC #11838
DOC Release dates for 0.20RC #11838
Conversation
|
If we want to do the logistic regression solver change I'd prefer to have it in the RC. |
|
I took a look at @amyxzhang original code for automatic extraction from an OPTICS ordered list, and the same bug from #11677 is present, so we haven't introduced it while modifying for the sklearn api and coding requirements. I wrote the optics graph builder and the dbscan_extract function, but I honestly don't have anymore insight into the auto_extract code then any other contributor that looks at it. From looking at the auto_extract outputs, at a very high level, I think you could describe the bug like so (although I'm guessing at this point): auto_extract builds a hierarchical clustering, and the root node contains all points (level 0), with each leaf being another cluster. For #11677, there are two clusters and two outliers. auto_cluster correctly sees the first cluster, and marks it as a child of root (level 1); then it processes the next cluster. Presumably it can't or won't move the two noise points up from level 1 to level 0 -- perhaps because they don't form an independent object (too few points), so every time they get moved up, they are pruned. I'm actually not sure why, but correct behavior would have a tree with three members-- two level 1 objects for each of the two clusters, and the level 0 root that contained the other two clusters and the noise points. |
|
@espg, could you copy the relevant part of the algorithm description paper to the issue so someone can check it line by line? |
|
any reason not to merge this? |
The idea would be to actually make the RC date specific, and branch, and create a what's new for 0.21 |
|
Today? ;) |
|
Maybe creating a PR at https://github.com/MacPython/scikit-learn-wheels and https://github.com/conda-forge/scikit-learn-feedstock for the RC and make sure that builds / tests succeeds there before creating the tag and branching out? |
|
I mean tagging is the last thing, we can branch though... Also maybe we want to start using scikit-ci at some point (not for this release though). |
|
We might need to make actual release September assuming there might be RC issues? |
|
Yeah, if the RC is released now, it doesn't seem likely that the final release will happen before September.
The CI in scikit-learn has so many customization, porting all of it to some third party system may take a lot of effort (and I wonder if it's worth it). I have not used scikit-ci, have used astropy-ci-helpers though: yes it makes 70% of things simpler for small projects, but for the other 30%, more time is wasted to make the system work around the way it was designed to work to adapt it to a particular project needs.. |
|
I like the idea of using scikit-ci but if I'm not mistaken it becomes less convenient when travis, appveyor and circle pipelines differs. And in scikit-learn we do different things in all CIs. I agree that it would be good to move from astropy-ci to scikit-ci. Eitherway we can ask @jcfr what are his thoughts. |
* tag '0.20rc1': (1109 commits) MNT rc version DOC Release dates for 0.20 (scikit-learn#11838) DOC Fix: require n_splits > 1 in TimeSeriesSplit (scikit-learn#11937) FIX xfail for MacOS LogisticRegressionCV stability (scikit-learn#11936) MNT: Use GEMV in enet_coordinate_descent (Pt. 1) (scikit-learn#11896) [MRG] TST/FIX stop optics reachability failure on 32bit (scikit-learn#11916) ENH add multi_class='auto' for LogisticRegression, default from 0.22; default solver will be 'lbfgs' (scikit-learn#11905) MAINT Fix test_logistic::test_dtype_match failure on 32 bit arch (scikit-learn#11899) DOC Updated link to Laurens van der Maaten's home page (scikit-learn#11907) DOC Remove stray backtick in /doc/modules/feature_extraction.rst (scikit-learn#11910) Deprecate min_samples_leaf and min_weight_fraction_leaf (scikit-learn#11870) MNT modify test_sparse_oneclasssvm to be parametrized (scikit-learn#11894) EXA set figure size to avoid overlaps (scikit-learn#11889) MRG/REL fixes /skips for 32bit tests (scikit-learn#11879) add durations=20 to makefile to show test runtimes locally (scikit-learn#11147) DOC loss='l2' is no longer accpeted in l1_min_c DOC add note about brute force nearest neighbors for string data (scikit-learn#11884) DOC Change sign of energy in RBM (scikit-learn#11156) RFC try to warn on iid less often (scikit-learn#11613) DOC reduce plot_gpr_prior_posterior.py warnings(scikit-learn#11664) ...
* releases: (1109 commits) MNT rc version DOC Release dates for 0.20 (scikit-learn#11838) DOC Fix: require n_splits > 1 in TimeSeriesSplit (scikit-learn#11937) FIX xfail for MacOS LogisticRegressionCV stability (scikit-learn#11936) MNT: Use GEMV in enet_coordinate_descent (Pt. 1) (scikit-learn#11896) [MRG] TST/FIX stop optics reachability failure on 32bit (scikit-learn#11916) ENH add multi_class='auto' for LogisticRegression, default from 0.22; default solver will be 'lbfgs' (scikit-learn#11905) MAINT Fix test_logistic::test_dtype_match failure on 32 bit arch (scikit-learn#11899) DOC Updated link to Laurens van der Maaten's home page (scikit-learn#11907) DOC Remove stray backtick in /doc/modules/feature_extraction.rst (scikit-learn#11910) Deprecate min_samples_leaf and min_weight_fraction_leaf (scikit-learn#11870) MNT modify test_sparse_oneclasssvm to be parametrized (scikit-learn#11894) EXA set figure size to avoid overlaps (scikit-learn#11889) MRG/REL fixes /skips for 32bit tests (scikit-learn#11879) add durations=20 to makefile to show test runtimes locally (scikit-learn#11147) DOC loss='l2' is no longer accpeted in l1_min_c DOC add note about brute force nearest neighbors for string data (scikit-learn#11884) DOC Change sign of energy in RBM (scikit-learn#11156) RFC try to warn on iid less often (scikit-learn#11613) DOC reduce plot_gpr_prior_posterior.py warnings(scikit-learn#11664) ...
* dfsg: (1109 commits) MNT rc version DOC Release dates for 0.20 (scikit-learn#11838) DOC Fix: require n_splits > 1 in TimeSeriesSplit (scikit-learn#11937) FIX xfail for MacOS LogisticRegressionCV stability (scikit-learn#11936) MNT: Use GEMV in enet_coordinate_descent (Pt. 1) (scikit-learn#11896) [MRG] TST/FIX stop optics reachability failure on 32bit (scikit-learn#11916) ENH add multi_class='auto' for LogisticRegression, default from 0.22; default solver will be 'lbfgs' (scikit-learn#11905) MAINT Fix test_logistic::test_dtype_match failure on 32 bit arch (scikit-learn#11899) DOC Updated link to Laurens van der Maaten's home page (scikit-learn#11907) DOC Remove stray backtick in /doc/modules/feature_extraction.rst (scikit-learn#11910) Deprecate min_samples_leaf and min_weight_fraction_leaf (scikit-learn#11870) MNT modify test_sparse_oneclasssvm to be parametrized (scikit-learn#11894) EXA set figure size to avoid overlaps (scikit-learn#11889) MRG/REL fixes /skips for 32bit tests (scikit-learn#11879) add durations=20 to makefile to show test runtimes locally (scikit-learn#11147) DOC loss='l2' is no longer accpeted in l1_min_c DOC add note about brute force nearest neighbors for string data (scikit-learn#11884) DOC Change sign of energy in RBM (scikit-learn#11156) RFC try to warn on iid less often (scikit-learn#11613) DOC reduce plot_gpr_prior_posterior.py warnings(scikit-learn#11664) ...
We are blocking on a fix to joblib/joblib#741. (See #11837) Otherwise, I think we can release an RC, with this change to release dates.
I'd be keen to add "Release Candidate 17 August 2018" if others are willing to go with this plan.
Note: This is merging to master, not 0.20.X. After merge we should branch to 0.20.X
Before release, we still need to add contributors to the docs. And perhaps we also need to sort out the OPTICS issue #11677 (@espg, any progress?).