From 207b07d3a9215211fe4c0c730406f665cf7dea77 Mon Sep 17 00:00:00 2001
From: Joel Nothman <joel.nothman@gmail.com>
Date: Thu, 22 Nov 2018 13:33:59 +1100
Subject: [PATCH 01/40] DOC add date for 0.20.0 release

---
 doc/whats_new/v0.20.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 4797c18217455..734faf1d8a9e0 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -215,7 +215,7 @@ Miscellaneous
 Version 0.20.0
 ==============
 
-**September, 2018**
+**September 25, 2018**
 
 This release packs in a mountain of bug fixes, features and enhancements for
 the Scikit-learn library, and improvements to the documentation and examples.

From 3794afa40293dd8c8cee456596b1263d0c07430a Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Thu, 22 Nov 2018 12:01:05 +0800
Subject: [PATCH 02/40] DOC update news, make 0.20 link to 0.20.0

---
 doc/index.rst | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/doc/index.rst b/doc/index.rst
index e5010099b245c..c665890474686 100644
--- a/doc/index.rst
+++ b/doc/index.rst
@@ -209,7 +209,9 @@
                     </li>
                     <li><strong>Scikit-learn 0.21 will drop support for Python 2.7 and Python 3.4.</strong>
                     </li>
-                    <li><em>September 2018.</em> scikit-learn 0.20 is available for download (<a href="whats_new.html#version-0-20">Changelog</a>).
+                    <li><em>November 2018.</em> scikit-learn 0.20.1 is available for download (<a href="whats_new.html#version-0-20-1">Changelog</a>).
+                    </li>
+                    <li><em>September 2018.</em> scikit-learn 0.20.0 is available for download (<a href="whats_new.html#version-0-20-0">Changelog</a>).
                     </li>
                     <li><em>July 2018.</em> scikit-learn 0.19.2 is available for download (<a href="whats_new.html#version-0-19">Changelog</a>).
                     </li>

From 4a4a73e23272c7446a2438a87eae3e44919fa094 Mon Sep 17 00:00:00 2001
From: Andreas Mueller <t3kcit@gmail.com>
Date: Thu, 22 Nov 2018 03:36:45 -0500
Subject: [PATCH 03/40] don't set check_pickle on new joblib on python2.7
 (#12645)

---
 sklearn/neighbors/base.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/sklearn/neighbors/base.py b/sklearn/neighbors/base.py
index f1efc0b6aeefc..16a52134098bc 100644
--- a/sklearn/neighbors/base.py
+++ b/sklearn/neighbors/base.py
@@ -439,11 +439,12 @@ class from an array representing our data set and ask who's
                 raise ValueError(
                     "%s does not work with sparse matrices. Densify the data, "
                     "or set algorithm='brute'" % self._fit_method)
-            if (sys.version_info < (3,) or
-                    LooseVersion(joblib_version) < LooseVersion('0.12')):
+            old_joblib = LooseVersion(joblib_version) < LooseVersion('0.12')
+            if sys.version_info < (3,) or old_joblib:
                 # Deal with change of API in joblib
+                check_pickle = False if old_joblib else None
                 delayed_query = delayed(_tree_query_parallel_helper,
-                                        check_pickle=False)
+                                        check_pickle=check_pickle)
                 parallel_kwargs = {"backend": "threading"}
             else:
                 delayed_query = delayed(_tree_query_parallel_helper)

From 677aa51e66ae8d7cef1ae9c255336673efcd1fc6 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Thu, 22 Nov 2018 23:03:37 +0800
Subject: [PATCH 04/40] DOC Add contributors for 0.20.1

---
 doc/whats_new/v0.20.rst | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 734faf1d8a9e0..3f46cbf50183d 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -210,6 +210,31 @@ Miscellaneous
   unvendoring joblib in future version of scikit-learn.
   :issue:`12345` by :user:`Thomas Moreau <tomMoral>`
 
+Code and Documentation Contributors
+-----------------------------------
+
+With thanks to:
+
+^__^, Adrin Jalali, Andrea Navarrete, Andreas Mueller,
+bauks, BenjaStudio, Cheuk Ting Ho, Connossor,
+Corey Levinson, Dan Stine, daten-kieker, Denis Kataev,
+Dillon Gardner, Dmitry Vukolov, Dougal J. Sutherland, Edward J Brown,
+Eric Chang, Federico Caselli, Gabriel Marzinotto, Gael Varoquaux,
+GauravAhlawat, Gustavo De Mari Pereira, Hanmin Qin, haroldfox,
+JackLangerman, Jacopo Notarstefano, janvanrijn, jdethurens,
+jeremiedbb, Joel Nothman, Joris Van den Bossche, Koen,
+Kushal Chauhan, Lee Yi Jie Joel, Lily Xiong, mail-liam,
+Mark Hannel, melsyt, Ming Li, Nicholas Smith,
+Nicolas Hug, Nikolay Shebanov, Oleksandr Pavlyk, Olivier Grisel,
+Peter Hausamann, Pierre Glaser, Pulkit Maloo, Quentin Batista,
+Radostin Stoyanov, Ramil Nugmanov, Rebekah Kim, Reshama Shaikh,
+Rohan Singh, Roman Feldbauer, Roman Yurchak, Roopam Sharma,
+Sam Waterbury, Scott Lowe, Sebastian Raschka, Stephen Tierney,
+SylvainLan, TakingItCasual, Thomas Fan, Thomas Moreau,
+Tom Dupré la Tour, Tulio Casagrande, Utkarsh Upadhyay, Xing Han Lu,
+Yaroslav Halchenko, Zach Miller
+
+
 .. _changes_0_20:
 
 Version 0.20.0

From e9866302a1b038b9ddce9e292a49ccb046d325a8 Mon Sep 17 00:00:00 2001
From: Paul Paczuski <pavopax@users.noreply.github.com>
Date: Thu, 22 Nov 2018 07:32:01 -0800
Subject: [PATCH 05/40] Doc: minor word change (#12646)

---
 doc/documentation.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/documentation.rst b/doc/documentation.rst
index 8ec2e41989a78..8a972de4c4f93 100644
--- a/doc/documentation.rst
+++ b/doc/documentation.rst
@@ -16,7 +16,7 @@ Documentation of scikit-learn |version|
                     <h2><a href="tutorial/basic/tutorial.html">Quick Start</a></h2>
                     <blockquote>A very short introduction into machine learning
                     problems and how to solve them using scikit-learn.
-                    Introduced basic concepts and conventions.
+                    Presents basic concepts and conventions.
                     </blockquote>
                 </div>
                 <div class="span4 box">

From df3058647e91b6c47cbbc4166b32119adbe32ffd Mon Sep 17 00:00:00 2001
From: Andreas Mueller <t3kcit@gmail.com>
Date: Thu, 22 Nov 2018 20:19:56 -0500
Subject: [PATCH 06/40] DOC fix minor sphinx issues (#12653)

---
 doc/testimonials/testimonials.rst | 8 ++++----
 doc/whats_new/v0.20.rst           | 4 ++--
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/doc/testimonials/testimonials.rst b/doc/testimonials/testimonials.rst
index 983b858f8ce96..0156df9e0abad 100644
--- a/doc/testimonials/testimonials.rst
+++ b/doc/testimonials/testimonials.rst
@@ -76,7 +76,7 @@ Erik Bernhardsson, Engineering Manager Music Discovery & Machine Learning, Spoti
    </span>
 
 `Inria <https://www.inria.fr/>`_
--------------------------------
+--------------------------------
 
 .. raw:: html
 
@@ -255,7 +255,7 @@ Alexandre Gramfort, Assistant Professor
 
 
 `Booking.com <https://www.booking.com>`_
--------------------------------------
+-----------------------------------------
 .. raw:: html
 
   <div class="logo">
@@ -728,7 +728,7 @@ Guillaume Lebourgeois & Samuel Charron - Data Scientists at Data Publica
 
 
 `Machinalis <https://www.machinalis.com/>`_
------------------------------------------
+-------------------------------------------
 
 .. raw:: html
 
@@ -937,7 +937,7 @@ Vlasios Vasileiou, Head of Data Science, Zopa
   </span>
 
 `MARS <https://www.mars.com/global>`_
-------------------------------------
+--------------------------------------
 
 .. raw:: html
 
diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 3f46cbf50183d..eedb8b645c443 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -94,7 +94,7 @@ Changelog
   :issue:`12388` by :user:`Connor Tann <Connossor>`.
 
 :mod:`sklearn.feature_extraction`
-...........................
+..................................
 
 - |Fix| Fixed a regression in v0.20.0 where
   :func:`feature_extraction.text.CountVectorizer` and other text vectorizers
@@ -152,7 +152,7 @@ Changelog
   :issue:`12171` by :user:`Thomas Moreau <tomMoral>`.
 
 :mod:`sklearn.preprocessing`
-........................
+.............................
 
 - |Fix| Fixed bug in :class:`preprocessing.OrdinalEncoder` when passing
   manually specified categories. :issue:`12365` by `Joris Van den Bossche`_.

From eea0a28fa7514626cd9897218d9a7bd2927bab7b Mon Sep 17 00:00:00 2001
From: Joris Van den Bossche <jorisvandenbossche@gmail.com>
Date: Fri, 23 Nov 2018 17:44:12 +0100
Subject: [PATCH 07/40] DOC: start section for the 0.20.2 bugfix notes (#12664)

---
 doc/whats_new/v0.20.rst | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index eedb8b645c443..629d34266cbb8 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -2,6 +2,22 @@
 
 .. currentmodule:: sklearn
 
+.. _changes_0_20_2:
+
+Version 0.20.2
+==============
+
+**XX, 2019**
+
+This is a bug-fix release with some minor documentation improvements and
+enhancements to features released in 0.20.0.
+
+Changelog
+---------
+
+
+
+
 .. _changes_0_20_1:
 
 Version 0.20.1

From 7e306a37a7d1494b6c293b22693c49e4747a0040 Mon Sep 17 00:00:00 2001
From: Albert Thomas <albertthomas88@gmail.com>
Date: Sat, 24 Nov 2018 09:49:10 +0100
Subject: [PATCH 08/40] DOC Fix links in anomaly detection example (#12665)

---
 examples/plot_anomaly_comparison.py | 44 +++++++++++++++--------------
 1 file changed, 23 insertions(+), 21 deletions(-)

diff --git a/examples/plot_anomaly_comparison.py b/examples/plot_anomaly_comparison.py
index f3dc0f1dddff1..203dcb6acd11b 100644
--- a/examples/plot_anomaly_comparison.py
+++ b/examples/plot_anomaly_comparison.py
@@ -14,32 +14,34 @@
 except for Local Outlier Factor (LOF) as it has no predict method to be applied
 on new data when it is used for outlier detection.
 
-The :class:`svm.OneClassSVM` is known to be sensitive to outliers and thus does
-not perform very well for outlier detection. This estimator is best suited for
-novelty detection when the training set is not contaminated by outliers.
-That said, outlier detection in high-dimension, or without any assumptions on
-the distribution of the inlying data is very challenging, and a One-class SVM
-might give useful results in these situations depending on the value of its
-hyperparameters.
-
-:class:`covariance.EllipticEnvelope` assumes the data is Gaussian and learns
-an ellipse. It thus degrades when the data is not unimodal. Notice however
-that this estimator is robust to outliers.
-
-:class:`ensemble.IsolationForest` and :class:`neighbors.LocalOutlierFactor`
-seem to perform reasonably well for multi-modal data sets. The advantage of
-:class:`neighbors.LocalOutlierFactor` over the other estimators is shown for
-the third data set, where the two modes have different densities. This
-advantage is explained by the local aspect of LOF, meaning that it only
+The :class:`sklearn.svm.OneClassSVM` is known to be sensitive to outliers and
+thus does not perform very well for outlier detection. This estimator is best
+suited for novelty detection when the training set is not contaminated by
+outliers. That said, outlier detection in high-dimension, or without any
+assumptions on the distribution of the inlying data is very challenging, and a
+One-class SVM might give useful results in these situations depending on the
+value of its hyperparameters.
+
+:class:`sklearn.covariance.EllipticEnvelope` assumes the data is Gaussian and
+learns an ellipse. It thus degrades when the data is not unimodal. Notice
+however that this estimator is robust to outliers.
+
+:class:`sklearn.ensemble.IsolationForest` and
+:class:`sklearn.neighbors.LocalOutlierFactor` seem to perform reasonably well
+for multi-modal data sets. The advantage of
+:class:`sklearn.neighbors.LocalOutlierFactor` over the other estimators is
+shown for the third data set, where the two modes have different densities.
+This advantage is explained by the local aspect of LOF, meaning that it only
 compares the score of abnormality of one sample with the scores of its
 neighbors.
 
 Finally, for the last data set, it is hard to say that one sample is more
 abnormal than another sample as they are uniformly distributed in a
-hypercube. Except for the :class:`svm.OneClassSVM` which overfits a little, all
-estimators present decent solutions for this situation. In such a case, it
-would be wise to look more closely at the scores of abnormality of the samples
-as a good estimator should assign similar scores to all the samples.
+hypercube. Except for the :class:`sklearn.svm.OneClassSVM` which overfits a
+little, all estimators present decent solutions for this situation. In such a
+case, it would be wise to look more closely at the scores of abnormality of
+the samples as a good estimator should assign similar scores to all the
+samples.
 
 While these examples give some intuition about the algorithms, this
 intuition might not apply to very high dimensional data.

From 498e45f37c53ee457360d8d90a503ba30e236131 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Lo=C3=AFc=20Est=C3=A8ve?= <loic.esteve@ymail.com>
Date: Sat, 24 Nov 2018 16:34:56 +0100
Subject: [PATCH 09/40] EXA remove unused variables. (#12647)

The data comes from static CSVs that cover 2003-01-01 to 2007-12-31.
---
 examples/applications/plot_stock_market.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/examples/applications/plot_stock_market.py b/examples/applications/plot_stock_market.py
index e2edc25b7eb72..6405f4d3e33b4 100644
--- a/examples/applications/plot_stock_market.py
+++ b/examples/applications/plot_stock_market.py
@@ -85,8 +85,6 @@
 # that we get high-tech firms, and before the 2008 crash). This kind of
 # historical data can be obtained for from APIs like the quandl.com and
 # alphavantage.co ones.
-start_date = datetime(2003, 1, 1).date()
-end_date = datetime(2008, 1, 1).date()
 
 symbol_dict = {
     'TOT': 'Total',

From 290492033161bbd9891b165538217f58338fe108 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Sat, 24 Nov 2018 23:38:38 +0800
Subject: [PATCH 10/40] MNT Unused import

---
 examples/applications/plot_stock_market.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/examples/applications/plot_stock_market.py b/examples/applications/plot_stock_market.py
index 6405f4d3e33b4..47d2eaa465d38 100644
--- a/examples/applications/plot_stock_market.py
+++ b/examples/applications/plot_stock_market.py
@@ -65,7 +65,6 @@
 # License: BSD 3 clause
 
 import sys
-from datetime import datetime
 
 import numpy as np
 import matplotlib.pyplot as plt

From 03ac3601ea34567078ca289efeba375a3f94d5c7 Mon Sep 17 00:00:00 2001
From: Nicolas Hug <contact@nicolas-hug.com>
Date: Sun, 25 Nov 2018 15:16:30 -0500
Subject: [PATCH 11/40] [MRG] Fixed randomness of test_logreg_cv_penalty
 (#12624)

* fixed randomness of test_logreg_cv_penalty

* Added comment about warm-starting coefficients
---
 sklearn/linear_model/tests/test_logistic.py | 29 ++++++++++++++++-----
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/sklearn/linear_model/tests/test_logistic.py b/sklearn/linear_model/tests/test_logistic.py
index fbc45dc40dfe9..b4ae8070154ed 100644
--- a/sklearn/linear_model/tests/test_logistic.py
+++ b/sklearn/linear_model/tests/test_logistic.py
@@ -1150,14 +1150,31 @@ def test_logreg_l1_sparse_data():
 
 @pytest.mark.filterwarnings('ignore: Default multi_class will')  # 0.22
 @pytest.mark.filterwarnings('ignore: You should specify a value')  # 0.22
-def test_logreg_cv_penalty():
-    # Test that the correct penalty is passed to the final fit.
-    X, y = make_classification(n_samples=50, n_features=20, random_state=0)
-    lr_cv = LogisticRegressionCV(penalty="l1", Cs=[1.0], solver='saga')
+@pytest.mark.parametrize("random_seed", [42])
+@pytest.mark.parametrize("penalty", ["l1", "l2"])
+def test_logistic_regression_cv_refit(random_seed, penalty):
+    # Test that when refit=True, logistic regression cv with the saga solver
+    # converges to the same solution as logistic regression with a fixed
+    # regularization parameter.
+    # Internally the LogisticRegressionCV model uses a warm start to refit on
+    # the full data model with the optimal C found by CV. As the penalized
+    # logistic regression loss is convex, we should still recover exactly
+    # the same solution as long as the stopping criterion is strict enough (and
+    # that there are no exactly duplicated features when penalty='l1').
+    X, y = make_classification(n_samples=50, n_features=20,
+                               random_state=random_seed)
+    common_params = dict(
+        solver='saga',
+        penalty=penalty,
+        random_state=random_seed,
+        max_iter=10000,
+        tol=1e-12,
+    )
+    lr_cv = LogisticRegressionCV(Cs=[1.0], refit=True, **common_params)
     lr_cv.fit(X, y)
-    lr = LogisticRegression(penalty="l1", C=1.0, solver='saga')
+    lr = LogisticRegression(C=1.0, **common_params)
     lr.fit(X, y)
-    assert_equal(np.count_nonzero(lr_cv.coef_), np.count_nonzero(lr.coef_))
+    assert_array_almost_equal(lr_cv.coef_, lr.coef_)
 
 
 def test_logreg_predict_proba_multinomial():

From a0005dafb94b9dc2de2f003a5077e99de38e0380 Mon Sep 17 00:00:00 2001
From: Sam Waterbury <samwaterbury1@gmail.com>
Date: Sun, 25 Nov 2018 15:29:00 -0600
Subject: [PATCH 12/40] DOC Updated MLP partial_fit documentation (#12618)

---
 sklearn/neural_network/multilayer_perceptron.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sklearn/neural_network/multilayer_perceptron.py b/sklearn/neural_network/multilayer_perceptron.py
index a8fcc8e11cdb9..f05c15c4ea2f3 100644
--- a/sklearn/neural_network/multilayer_perceptron.py
+++ b/sklearn/neural_network/multilayer_perceptron.py
@@ -619,7 +619,7 @@ def fit(self, X, y):
 
     @property
     def partial_fit(self):
-        """Fit the model to data matrix X and target y.
+        """Update the model with a single iteration over the given data.
 
         Parameters
         ----------
@@ -978,7 +978,7 @@ def fit(self, X, y):
 
     @property
     def partial_fit(self):
-        """Fit the model to data matrix X and target y.
+        """Update the model with a single iteration over the given data.
 
         Parameters
         ----------

From c24e63e070ceb76fb8b7cb8d0958a7cbacf73434 Mon Sep 17 00:00:00 2001
From: Olivier Grisel <olivier.grisel@ensta.org>
Date: Mon, 26 Nov 2018 05:18:14 +0100
Subject: [PATCH 13/40] DOC use twine to upload wheels to pypi (#12661)

---
 doc/developers/maintainer.rst | 30 +++++++++++++++++++++---------
 1 file changed, 21 insertions(+), 9 deletions(-)

diff --git a/doc/developers/maintainer.rst b/doc/developers/maintainer.rst
index a3309abcfbf10..107c1c0620c60 100644
--- a/doc/developers/maintainer.rst
+++ b/doc/developers/maintainer.rst
@@ -33,21 +33,21 @@ For more information see https://github.com/scikit-learn/scikit-learn/wiki/How-t
 
     $ git push origin --tags
 
-4. create tarballs:
+4. create the source tarball:
 
    - Wipe clean your repo::
 
        $ git clean -xfd
 
-   - Register and upload on PyPI::
+   - Generate the tarball::
 
-       $ python setup.py sdist register upload
+       $ python setup.py sdist
 
+   The result should be in the `dist/` folder. We will upload it later
+   with the wheels. Check that you can install it in a new virtualenv and
+   that the tests pass.
 
-5. Push the documentation to the website. Circle CI should do this
-   automatically for master and <N>.<N>.X branches.
-
-6. Build binaries using dedicated CI servers by updating the git submodule
+5. Build binaries using dedicated CI servers by updating the git submodule
    reference to the new scikit-learn tag of the release at:
 
    https://github.com/MacPython/scikit-learn-wheels
@@ -56,9 +56,21 @@ For more information see https://github.com/scikit-learn/scikit-learn/wiki/How-t
    packages and upload them to PyPI by running the following commands in the
    scikit-learn source folder (checked out at the release tag)::
 
-       $ pip install -U wheelhouse_uploader
-       $ python setup.py sdist fetch_artifacts upload_all
+       $ pip install -U wheelhouse_uploader twine
+       $ python setup.py fetch_artifacts
+
+   Check the content of the `dist/` folder: it should contain all the wheels
+   along with the source tarball ("scikit-learn-XXX.tar.gz").
+
+   Make sure that you do not have developer versions or older versions of
+   the scikit-learn package in that folder.
 
+   Upload everything at once to https://pypi.org::
+
+       $ twine upload dist/
+
+6. Push the documentation to the website. Circle CI should do this
+   automatically for master and <N>.<N>.X branches.
 
 7. FOR FINAL RELEASE: Update the release date in What's New
 

From 039f3653815b873259a3204119e7f7157cd7bd7e Mon Sep 17 00:00:00 2001
From: josephsalmon <josephsalmon@users.noreply.github.com>
Date: Tue, 27 Nov 2018 21:07:38 +0100
Subject: [PATCH 14/40] EXA improving examples/mixture/plot_gmm_covariances.py
 visualisation (#12683)

---
 examples/mixture/plot_gmm_covariances.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/examples/mixture/plot_gmm_covariances.py b/examples/mixture/plot_gmm_covariances.py
index 2b4cd88642a98..14a0efccc01da 100644
--- a/examples/mixture/plot_gmm_covariances.py
+++ b/examples/mixture/plot_gmm_covariances.py
@@ -64,6 +64,7 @@ def make_ellipses(gmm, ax):
         ell.set_clip_box(ax.bbox)
         ell.set_alpha(0.5)
         ax.add_artist(ell)
+        ax.set_aspect('equal', 'datalim')
 
 iris = datasets.load_iris()
 

From 6c1703f03f3b71f51b4c2043ee663d7b7228fc7e Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Wed, 28 Nov 2018 04:17:49 +0800
Subject: [PATCH 15/40] TST Specify random seed in
 test_shuffle_kfold_stratifiedkfold_reproducibility (#12674)

---
 sklearn/model_selection/tests/test_split.py | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/sklearn/model_selection/tests/test_split.py b/sklearn/model_selection/tests/test_split.py
index 308c94a2e273b..887a6b0a315c8 100644
--- a/sklearn/model_selection/tests/test_split.py
+++ b/sklearn/model_selection/tests/test_split.py
@@ -458,13 +458,13 @@ def test_shuffle_kfold():
 
 
 def test_shuffle_kfold_stratifiedkfold_reproducibility():
-    # Check that when the shuffle is True multiple split calls produce the
-    # same split when random_state is set
     X = np.ones(15)  # Divisible by 3
     y = [0] * 7 + [1] * 8
     X2 = np.ones(16)  # Not divisible by 3
     y2 = [0] * 8 + [1] * 8
 
+    # Check that when the shuffle is True, multiple split calls produce the
+    # same split when random_state is int
     kf = KFold(3, shuffle=True, random_state=0)
     skf = StratifiedKFold(3, shuffle=True, random_state=0)
 
@@ -472,8 +472,12 @@ def test_shuffle_kfold_stratifiedkfold_reproducibility():
         np.testing.assert_equal(list(cv.split(X, y)), list(cv.split(X, y)))
         np.testing.assert_equal(list(cv.split(X2, y2)), list(cv.split(X2, y2)))
 
-    kf = KFold(3, shuffle=True)
-    skf = StratifiedKFold(3, shuffle=True)
+    # Check that when the shuffle is True, multiple split calls often
+    # (not always) produce different splits when random_state is
+    # RandomState instance or None
+    kf = KFold(3, shuffle=True, random_state=np.random.RandomState(0))
+    skf = StratifiedKFold(3, shuffle=True,
+                          random_state=np.random.RandomState(0))
 
     for cv in (kf, skf):
         for data in zip((X, X2), (y, y2)):

From 676bf9bc4eb77b4f10eb94c6b9f5ecc0ce055b68 Mon Sep 17 00:00:00 2001
From: Adrin Jalali <adrin.jalali@gmail.com>
Date: Wed, 28 Nov 2018 23:40:28 +0100
Subject: [PATCH 16/40] DOC add joblib to doc-build deps (#12691)

---
 doc/developers/contributing.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/developers/contributing.rst b/doc/developers/contributing.rst
index 0bb8b1fc39f59..0251638a71378 100644
--- a/doc/developers/contributing.rst
+++ b/doc/developers/contributing.rst
@@ -432,7 +432,7 @@ Building the documentation
 
 Building the documentation requires installing some additional packages::
 
-    pip install sphinx sphinx-gallery numpydoc matplotlib Pillow pandas scikit-image
+    pip install sphinx sphinx-gallery numpydoc matplotlib Pillow pandas scikit-image joblib
 
 To build the documentation, you need to be in the ``doc`` folder::
 

From caf652f23952878eaebad9dd51b0e444908bbfcc Mon Sep 17 00:00:00 2001
From: Thomas Fan <thomasjpfan@gmail.com>
Date: Thu, 29 Nov 2018 20:00:44 -0500
Subject: [PATCH 17/40] FIX Adapts cdist jaccard to scipy 1.2.0 (#12692)

---
 doc/whats_new/v0.20.rst                      | 16 ++++++++++++++++
 sklearn/neighbors/dist_metrics.pyx           |  5 +++++
 sklearn/neighbors/tests/test_dist_metrics.py |  7 +++++++
 3 files changed, 28 insertions(+)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 629d34266cbb8..0316a59668277 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -12,11 +12,27 @@ Version 0.20.2
 This is a bug-fix release with some minor documentation improvements and
 enhancements to features released in 0.20.0.
 
+Changed models
+--------------
+
+The following estimators and functions, when fit with the same data and
+parameters, may produce different models from the previous version. This often
+occurs due to changes in the modelling logic (bug fixes or enhancements), or in
+random sampling procedures.
+
+- :mod:`sklearn.neighbors` when ``metric=='jaccard'`` (bug fix)
+
 Changelog
 ---------
 
 
 
+:mod:`sklearn.neighbors`
+........................
+
+- |Fix| Fixed :class:`sklearn.neighbors.DistanceMetric` jaccard distance
+  function to return 0 when two all-zero vectors are compared.
+  :issue:`12685` by :user:`Thomas Fan <thomasjpfan>`.
 
 .. _changes_0_20_1:
 
diff --git a/sklearn/neighbors/dist_metrics.pyx b/sklearn/neighbors/dist_metrics.pyx
index 440187d60e49f..ca546c1bd245d 100755
--- a/sklearn/neighbors/dist_metrics.pyx
+++ b/sklearn/neighbors/dist_metrics.pyx
@@ -788,6 +788,11 @@ cdef class JaccardDistance(DistanceMetric):
             tf2 = x2[j] != 0
             nnz += (tf1 or tf2)
             n_eq += (tf1 and tf2)
+        # Based on https://github.com/scipy/scipy/pull/7373
+        # When comparing two all-zero vectors, scipy>=1.2.0 jaccard metric
+        # was changed to return 0, instead of nan.
+        if nnz == 0:
+            return 0
         return (nnz - n_eq) * 1.0 / nnz
 
 
diff --git a/sklearn/neighbors/tests/test_dist_metrics.py b/sklearn/neighbors/tests/test_dist_metrics.py
index f4d6dc3e74c5e..67be5631d715f 100644
--- a/sklearn/neighbors/tests/test_dist_metrics.py
+++ b/sklearn/neighbors/tests/test_dist_metrics.py
@@ -6,6 +6,8 @@
 
 import pytest
 
+from distutils.version import LooseVersion
+from scipy import __version__ as scipy_version
 from scipy.spatial.distance import cdist
 from sklearn.neighbors.dist_metrics import DistanceMetric
 from sklearn.neighbors import BallTree
@@ -101,6 +103,11 @@ def check_pdist(metric, kwargs, D_true):
 def check_pdist_bool(metric, D_true):
     dm = DistanceMetric.get_metric(metric)
     D12 = dm.pairwise(X1_bool)
+    # Based on https://github.com/scipy/scipy/pull/7373
+    # When comparing two all-zero vectors, scipy>=1.2.0 jaccard metric
+    # was changed to return 0, instead of nan.
+    if metric == 'jaccard' and LooseVersion(scipy_version) < '1.2.0':
+        D_true[np.isnan(D_true)] = 0
     assert_array_almost_equal(D12, D_true)
 
 

From 4822a19c2fc3125f9b536eda9c4ea3d5a325e5ad Mon Sep 17 00:00:00 2001
From: Shivam Kotwalia <shivamkotwalia@gmail.com>
Date: Fri, 30 Nov 2018 17:46:57 +0530
Subject: [PATCH 18/40] DOC Fix joblib URL (#12698)

---
 doc/developers/performance.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/developers/performance.rst b/doc/developers/performance.rst
index 325199a464fab..861cb78fdf78b 100644
--- a/doc/developers/performance.rst
+++ b/doc/developers/performance.rst
@@ -383,7 +383,7 @@ TODO: give a simple teaser example here.
 
 Checkout the official joblib documentation:
 
-- https://pythonhosted.org/joblib
+- https://joblib.readthedocs.io
 
 
 .. _warm-restarts:

From dc5359fc540161ccca40a60aff95348c325a2834 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Fri, 30 Nov 2018 23:19:29 +0800
Subject: [PATCH 19/40] MNT Minor code refactoring in utils.testing (#12688)

---
 sklearn/utils/testing.py | 33 +++++++++++++--------------------
 1 file changed, 13 insertions(+), 20 deletions(-)

diff --git a/sklearn/utils/testing.py b/sklearn/utils/testing.py
index db43cce6fbaf9..e5fd6146740c0 100644
--- a/sklearn/utils/testing.py
+++ b/sklearn/utils/testing.py
@@ -42,12 +42,21 @@
 except NameError:
     WindowsError = None
 
+from numpy.testing import assert_allclose
+from numpy.testing import assert_almost_equal
+from numpy.testing import assert_approx_equal
+from numpy.testing import assert_array_equal
+from numpy.testing import assert_array_almost_equal
+from numpy.testing import assert_array_less
+import numpy as np
+
 import sklearn
-from sklearn.base import BaseEstimator
+from sklearn.base import (BaseEstimator, ClassifierMixin, ClusterMixin,
+                          RegressorMixin, TransformerMixin)
+from sklearn.utils import deprecated, IS_PYPY, _IS_32BIT
 from sklearn.utils._joblib import joblib
+from sklearn.utils._unittest_backport import TestCase
 from sklearn.utils.fixes import signature
-from sklearn.utils import deprecated, IS_PYPY, _IS_32BIT
-
 
 additional_names_in_all = []
 try:
@@ -73,24 +82,13 @@
 except ImportError:
     pass
 
-from numpy.testing import assert_almost_equal
-from numpy.testing import assert_array_equal
-from numpy.testing import assert_array_almost_equal
-from numpy.testing import assert_array_less
-from numpy.testing import assert_approx_equal
-import numpy as np
-
-from sklearn.base import (ClassifierMixin, RegressorMixin, TransformerMixin,
-                          ClusterMixin)
-from sklearn.utils._unittest_backport import TestCase
-
 __all__ = ["assert_equal", "assert_not_equal", "assert_raises",
            "assert_raises_regexp", "assert_true",
            "assert_false", "assert_almost_equal", "assert_array_equal",
            "assert_array_almost_equal", "assert_array_less",
            "assert_less", "assert_less_equal",
            "assert_greater", "assert_greater_equal",
-           "assert_approx_equal", "SkipTest"]
+           "assert_approx_equal", "assert_allclose", "SkipTest"]
 __all__.extend(additional_names_in_all)
 
 _dummy = TestCase('__init__')
@@ -379,11 +377,6 @@ def __exit__(self, *exc_info):
         clean_warning_registry()
 
 
-assert_less = _dummy.assertLess
-assert_greater = _dummy.assertGreater
-
-assert_allclose = np.testing.assert_allclose
-
 def assert_raise_message(exceptions, message, function, *args, **kwargs):
     """Helper function to test the message raised in an exception.
 

From e9f0c1db7afa5f5c0f7ad97a793ca9faee0a3a15 Mon Sep 17 00:00:00 2001
From: Dan Stine <dstine@users.noreply.github.com>
Date: Sun, 2 Dec 2018 20:21:02 -0500
Subject: [PATCH 20/40] DOC Update TfidfVectorizer analyzer documentation to
 match CountVectorizer (#12713)

---
 sklearn/feature_extraction/text.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/sklearn/feature_extraction/text.py b/sklearn/feature_extraction/text.py
index 82d5c6614d310..a5a25d8208e2a 100644
--- a/sklearn/feature_extraction/text.py
+++ b/sklearn/feature_extraction/text.py
@@ -1356,8 +1356,10 @@ class TfidfVectorizer(CountVectorizer):
         preprocessing and n-grams generation steps.
         Only applies if ``analyzer == 'word'``.
 
-    analyzer : string, {'word', 'char'} or callable
+    analyzer : string, {'word', 'char', 'char_wb'} or callable
         Whether the feature should be made of word or character n-grams.
+        Option 'char_wb' creates character n-grams only from text inside
+        word boundaries; n-grams at the edges of words are padded with space.
 
         If a callable is passed it is used to extract the sequence of features
         out of the raw, unprocessed input.

From b1471d21c861cf32a513b35293af84c185eb0f9e Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Mon, 3 Dec 2018 18:22:32 +0800
Subject: [PATCH 21/40] FIX Fix error in make_column_transformer when columns
 is pandas.Index (#12704)

Fixes #12703
---
 doc/whats_new/v0.20.rst                          |  6 ++++++
 sklearn/compose/_column_transformer.py           |  2 +-
 sklearn/compose/tests/test_column_transformer.py | 14 ++++++++++++++
 3 files changed, 21 insertions(+), 1 deletion(-)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 0316a59668277..aca820ac55ff6 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -25,6 +25,12 @@ random sampling procedures.
 Changelog
 ---------
 
+:mod:`sklearn.compose`
+......................
+
+- |Fix| Fixed an issue in :func:`compose.make_column_transformer` which raises
+  unexpected error when columns is pandas Index or pandas Series.
+  :issue:`12704` by :user:`Hanmin Qin <qinhanmin2014>`.
 
 
 :mod:`sklearn.neighbors`
diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 088a7f44c1bda..1e020cb95068c 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -692,7 +692,7 @@ def _validate_transformers(transformers):
         return True
 
     for t in transformers:
-        if t in ('drop', 'passthrough'):
+        if isinstance(t, six.string_types) and t in ('drop', 'passthrough'):
             continue
         if (not (hasattr(t, "fit") or hasattr(t, "fit_transform")) or not
                 hasattr(t, "transform")):
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index c201cfef23daa..d564aa097a63a 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -542,6 +542,20 @@ def test_make_column_transformer():
                                 ('first', 'drop'))
 
 
+def test_make_column_transformer_pandas():
+    pd = pytest.importorskip('pandas')
+    X_array = np.array([[0, 1, 2], [2, 4, 6]]).T
+    X_df = pd.DataFrame(X_array, columns=['first', 'second'])
+    norm = Normalizer()
+    # XXX remove in v0.22
+    with pytest.warns(DeprecationWarning,
+                      match='`make_column_transformer` now expects'):
+        ct1 = make_column_transformer((X_df.columns, norm))
+    ct2 = make_column_transformer((norm, X_df.columns))
+    assert_almost_equal(ct1.fit_transform(X_df),
+                        ct2.fit_transform(X_df))
+
+
 def test_make_column_transformer_kwargs():
     scaler = StandardScaler()
     norm = Normalizer()

From 765f3e46f61a842afd7bfe827640ea622cf5b8ae Mon Sep 17 00:00:00 2001
From: Joris Van den Bossche <jorisvandenbossche@gmail.com>
Date: Mon, 3 Dec 2018 11:25:52 +0100
Subject: [PATCH 22/40] BUG: fix check_array on pandas Series with custom dtype
 (eg categorical) (#12706)

Closes #12699. Related to #12625
---
 doc/whats_new/v0.20.rst                | 7 +++++++
 sklearn/utils/tests/test_validation.py | 5 +++++
 sklearn/utils/validation.py            | 2 +-
 3 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index aca820ac55ff6..4581d500ab307 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -40,6 +40,13 @@ Changelog
   function to return 0 when two all-zero vectors are compared.
   :issue:`12685` by :user:`Thomas Fan <thomasjpfan>`.
 
+:mod:`sklearn.utils`
+....................
+
+- |Fix| Calling :func:`utils.check_array` on `pandas.Series` with categorical
+  data, which raised an error in 0.20.0, now returns the expected output again.
+  :issue:`12699` by `Joris Van den Bossche`_.
+
 .. _changes_0_20_1:
 
 Version 0.20.1
diff --git a/sklearn/utils/tests/test_validation.py b/sklearn/utils/tests/test_validation.py
index 9ff3794ff38ff..b3af8e3a914ab 100644
--- a/sklearn/utils/tests/test_validation.py
+++ b/sklearn/utils/tests/test_validation.py
@@ -700,6 +700,11 @@ def test_check_array_series():
                       warn_on_dtype=True)
     assert_array_equal(res, np.array([1, 2, 3]))
 
+    # with categorical dtype (not a numpy dtype) (GH12699)
+    s = pd.Series(['a', 'b', 'c']).astype('category')
+    res = check_array(s, dtype=None, ensure_2d=False)
+    assert_array_equal(res, np.array(['a', 'b', 'c'], dtype=object))
+
 
 def test_check_dataframe_warns_on_dtype():
     # Check that warn_on_dtype also works for DataFrames.
diff --git a/sklearn/utils/validation.py b/sklearn/utils/validation.py
index 305e9c4d30bf6..be35f47429af5 100644
--- a/sklearn/utils/validation.py
+++ b/sklearn/utils/validation.py
@@ -477,7 +477,7 @@ def check_array(array, accept_sparse=False, accept_large_sparse=True,
     # check if the object contains several dtypes (typically a pandas
     # DataFrame), and store them. If not, store None.
     dtypes_orig = None
-    if hasattr(array, "dtypes") and len(array.dtypes):
+    if hasattr(array, "dtypes") and hasattr(array.dtypes, '__array__'):
         dtypes_orig = np.array(array.dtypes)
 
     if dtype_numeric:

From 3af7147ffb90308acf2ccd6ee9fd6476ae090e14 Mon Sep 17 00:00:00 2001
From: Joel Nothman <joel.nothman@gmail.com>
Date: Wed, 5 Dec 2018 09:58:34 +1100
Subject: [PATCH 23/40] DOC regenerate authors.rst after Adrin has joined
 (#12719)

---
 build_tools/generate_authors_table.py |  2 +-
 doc/authors.rst                       | 26 ++++++++++++++++----------
 2 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/build_tools/generate_authors_table.py b/build_tools/generate_authors_table.py
index ea3796473396d..4d0a2349e1cdb 100644
--- a/build_tools/generate_authors_table.py
+++ b/build_tools/generate_authors_table.py
@@ -97,7 +97,7 @@ def key(profile):
 contributors = get_contributors()
 
 print(".. raw :: html\n")
-print("    <!-- Generated by gen_authors.py -->")
+print("    <!-- Generated by generate_authors_table.py -->")
 print("    <table>")
 print("    <col style='width:%d%%' span='%d'>"
       % (int(100 / ROW_SIZE), ROW_SIZE))
diff --git a/doc/authors.rst b/doc/authors.rst
index 0210dff4bef6e..160cbbb3512b0 100644
--- a/doc/authors.rst
+++ b/doc/authors.rst
@@ -1,6 +1,6 @@
 .. raw :: html
 
-    <!-- Generated by gen_authors.py -->
+    <!-- Generated by generate_authors_table.py -->
     <table>
     <col style='width:14%' span='7'>
     <style>
@@ -70,7 +70,7 @@
     <tr>
     <td>
     <a href='https://github.com/chrisfilo'><img src='https://avatars2.githubusercontent.com/u/238759?v=4' class='avatar' /></a> <br />
-    <p>Chris Filo Gorgolewski</p>
+    <p>Chris Gorgolewski</p>
     </td>
     <td>
     <a href='https://github.com/agramfort'><img src='https://avatars2.githubusercontent.com/u/161052?v=4' class='avatar' /></a> <br />
@@ -93,12 +93,16 @@
     <p>Brian Holt</p>
     </td>
     <td>
-    <a href='https://github.com/arjoly'><img src='https://avatars0.githubusercontent.com/u/1274722?v=4' class='avatar' /></a> <br />
-    <p>Arnaud Joly</p>
+    <a href='https://github.com/adrinjalali'><img src='https://avatars3.githubusercontent.com/u/1663864?v=4' class='avatar' /></a> <br />
+    <p>Adrin Jalali</p>
     </td>
     </tr>
     <tr>
     <td>
+    <a href='https://github.com/arjoly'><img src='https://avatars0.githubusercontent.com/u/1274722?v=4' class='avatar' /></a> <br />
+    <p>Arnaud Joly</p>
+    </td>
+    <td>
     <a href='https://github.com/thouis'><img src='https://avatars1.githubusercontent.com/u/473043?v=4' class='avatar' /></a> <br />
     <p>Thouis (Ray) Jones</p>
     </td>
@@ -122,12 +126,12 @@
     <a href='https://github.com/weilinear'><img src='https://avatars0.githubusercontent.com/u/2232328?v=4' class='avatar' /></a> <br />
     <p>Wei Li</p>
     </td>
+    </tr>
+    <tr>
     <td>
     <a href='https://github.com/paolo-losi'><img src='https://avatars1.githubusercontent.com/u/264906?v=4' class='avatar' /></a> <br />
     <p>Paolo Losi</p>
     </td>
-    </tr>
-    <tr>
     <td>
     <a href='https://github.com/glouppe'><img src='https://avatars3.githubusercontent.com/u/477771?v=4' class='avatar' /></a> <br />
     <p>Gilles Louppe</p>
@@ -152,12 +156,12 @@
     <a href='https://github.com/vene'><img src='https://avatars0.githubusercontent.com/u/241745?v=4' class='avatar' /></a> <br />
     <p>Vlad Niculae</p>
     </td>
+    </tr>
+    <tr>
     <td>
     <a href='https://github.com/jnothman'><img src='https://avatars2.githubusercontent.com/u/78827?v=4' class='avatar' /></a> <br />
     <p>Joel Nothman</p>
     </td>
-    </tr>
-    <tr>
     <td>
     <a href='https://github.com/alextp'><img src='https://avatars0.githubusercontent.com/u/5061?v=4' class='avatar' /></a> <br />
     <p>Alexandre Passos</p>
@@ -182,12 +186,12 @@
     <a href='https://github.com/jmschrei'><img src='https://avatars2.githubusercontent.com/u/3916816?v=4' class='avatar' /></a> <br />
     <p>Jacob Schreiber</p>
     </td>
+    </tr>
+    <tr>
     <td>
     <a href='https://github.com/bthirion'><img src='https://avatars1.githubusercontent.com/u/234454?v=4' class='avatar' /></a> <br />
     <p>Bertrand Thirion</p>
     </td>
-    </tr>
-    <tr>
     <td>
     <a href='https://github.com/TomDLT'><img src='https://avatars2.githubusercontent.com/u/11065596?v=4' class='avatar' /></a> <br />
     <p>Tom Dupré la Tour</p>
@@ -212,6 +216,8 @@
     <a href='https://github.com/ronw'><img src='https://avatars2.githubusercontent.com/u/113819?v=4' class='avatar' /></a> <br />
     <p>Ron Weiss</p>
     </td>
+    </tr>
+    <tr>
     <td>
     <a href='https://github.com/rth'><img src='https://avatars0.githubusercontent.com/u/630936?v=4' class='avatar' /></a> <br />
     <p>Roman Yurchak</p>

From 9d9d798333b2b61a55fe62faa3fc29d716d26b89 Mon Sep 17 00:00:00 2001
From: Adrin Jalali <adrin.jalali@gmail.com>
Date: Thu, 6 Dec 2018 02:05:41 +0100
Subject: [PATCH 24/40] EXA change normed to density in matplotlib calls in
 examples (#12718)

---
 examples/compose/plot_transformed_target.py  | 17 +++++++++++++----
 examples/neighbors/plot_kde_1d.py            | 11 +++++++++--
 examples/plot_johnson_lindenstrauss_bound.py | 10 +++++++++-
 3 files changed, 31 insertions(+), 7 deletions(-)

diff --git a/examples/compose/plot_transformed_target.py b/examples/compose/plot_transformed_target.py
index ae4d6e6981032..d9bb0c7f5b091 100755
--- a/examples/compose/plot_transformed_target.py
+++ b/examples/compose/plot_transformed_target.py
@@ -20,7 +20,9 @@
 from __future__ import print_function, division
 
 import numpy as np
+import matplotlib
 import matplotlib.pyplot as plt
+from distutils.version import LooseVersion
 
 print(__doc__)
 
@@ -34,6 +36,13 @@
 from sklearn.compose import TransformedTargetRegressor
 from sklearn.metrics import median_absolute_error, r2_score
 
+
+# `normed` is being deprecated in favor of `density` in histograms
+if LooseVersion(matplotlib.__version__) >= '2.1':
+    density_param = {'density': True}
+else:
+    density_param = {'normed': True}
+
 ###############################################################################
 # A synthetic random regression problem is generated. The targets ``y`` are
 # modified by: (i) translating all targets such that all entries are
@@ -54,13 +63,13 @@
 
 f, (ax0, ax1) = plt.subplots(1, 2)
 
-ax0.hist(y, bins=100, normed=True)
+ax0.hist(y, bins=100, **density_param)
 ax0.set_xlim([0, 2000])
 ax0.set_ylabel('Probability')
 ax0.set_xlabel('Target')
 ax0.set_title('Target distribution')
 
-ax1.hist(y_trans, bins=100, normed=True)
+ax1.hist(y_trans, bins=100, **density_param)
 ax1.set_ylabel('Probability')
 ax1.set_xlabel('Target')
 ax1.set_title('Transformed target distribution')
@@ -139,12 +148,12 @@
 
 f, (ax0, ax1) = plt.subplots(1, 2)
 
-ax0.hist(y, bins=100, normed=True)
+ax0.hist(y, bins=100, **density_param)
 ax0.set_ylabel('Probability')
 ax0.set_xlabel('Target')
 ax0.set_title('Target distribution')
 
-ax1.hist(y_trans, bins=100, normed=True)
+ax1.hist(y_trans, bins=100, **density_param)
 ax1.set_ylabel('Probability')
 ax1.set_xlabel('Target')
 ax1.set_title('Transformed target distribution')
diff --git a/examples/neighbors/plot_kde_1d.py b/examples/neighbors/plot_kde_1d.py
index aac5e6ea3e7fa..8e1673dd48944 100644
--- a/examples/neighbors/plot_kde_1d.py
+++ b/examples/neighbors/plot_kde_1d.py
@@ -29,10 +29,17 @@
 # Author: Jake Vanderplas <jakevdp@cs.washington.edu>
 #
 import numpy as np
+import matplotlib
 import matplotlib.pyplot as plt
+from distutils.version import LooseVersion
 from scipy.stats import norm
 from sklearn.neighbors import KernelDensity
 
+# `normed` is being deprecated in favor of `density` in histograms
+if LooseVersion(matplotlib.__version__) >= '2.1':
+    density_param = {'density': True}
+else:
+    density_param = {'normed': True}
 
 #----------------------------------------------------------------------
 # Plot the progression of histograms to kernels
@@ -47,11 +54,11 @@
 fig.subplots_adjust(hspace=0.05, wspace=0.05)
 
 # histogram 1
-ax[0, 0].hist(X[:, 0], bins=bins, fc='#AAAAFF', normed=True)
+ax[0, 0].hist(X[:, 0], bins=bins, fc='#AAAAFF', **density_param)
 ax[0, 0].text(-3.5, 0.31, "Histogram")
 
 # histogram 2
-ax[0, 1].hist(X[:, 0], bins=bins + 0.75, fc='#AAAAFF', normed=True)
+ax[0, 1].hist(X[:, 0], bins=bins + 0.75, fc='#AAAAFF', **density_param)
 ax[0, 1].text(-3.5, 0.31, "Histogram, bins shifted")
 
 # tophat KDE
diff --git a/examples/plot_johnson_lindenstrauss_bound.py b/examples/plot_johnson_lindenstrauss_bound.py
index cc711f48d52c6..9be19045e8463 100644
--- a/examples/plot_johnson_lindenstrauss_bound.py
+++ b/examples/plot_johnson_lindenstrauss_bound.py
@@ -94,13 +94,21 @@
 import sys
 from time import time
 import numpy as np
+import matplotlib
 import matplotlib.pyplot as plt
+from distutils.version import LooseVersion
 from sklearn.random_projection import johnson_lindenstrauss_min_dim
 from sklearn.random_projection import SparseRandomProjection
 from sklearn.datasets import fetch_20newsgroups_vectorized
 from sklearn.datasets import load_digits
 from sklearn.metrics.pairwise import euclidean_distances
 
+# `normed` is being deprecated in favor of `density` in histograms
+if LooseVersion(matplotlib.__version__) >= '2.1':
+    density_param = {'density': True}
+else:
+    density_param = {'normed': True}
+
 # Part 1: plot the theoretical dependency between n_components_min and
 # n_samples
 
@@ -187,7 +195,7 @@
           % (np.mean(rates), np.std(rates)))
 
     plt.figure()
-    plt.hist(rates, bins=50, normed=True, range=(0., 2.), edgecolor='k')
+    plt.hist(rates, bins=50, range=(0., 2.), edgecolor='k', **density_param)
     plt.xlabel("Squared distances rate: projected / original")
     plt.ylabel("Distribution of samples pairs")
     plt.title("Histogram of pairwise distance rates for n_components=%d" %

From fdfba9d0bd39cacc9534190523e17fbc0f4dd75c Mon Sep 17 00:00:00 2001
From: Jan S <github_new@milania.de>
Date: Mon, 10 Dec 2018 11:33:07 +0100
Subject: [PATCH 25/40] DOC Clarify what the decision function in SVM
 calculates (#12708)

---
 sklearn/svm/base.py | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/sklearn/svm/base.py b/sklearn/svm/base.py
index 248ae196d9400..77dc9d9283f67 100644
--- a/sklearn/svm/base.py
+++ b/sklearn/svm/base.py
@@ -384,7 +384,7 @@ def _compute_kernel(self, X):
         return X
 
     def _decision_function(self, X):
-        """Distance of the samples X to the separating hyperplane.
+        """Evaluates the decision function for the samples in X.
 
         Parameters
         ----------
@@ -529,7 +529,7 @@ def _validate_targets(self, y):
         return np.asarray(y, dtype=np.float64, order='C')
 
     def decision_function(self, X):
-        """Distance of the samples X to the separating hyperplane.
+        """Evaluates the decision function for the samples in X.
 
         Parameters
         ----------
@@ -541,7 +541,16 @@ def decision_function(self, X):
             Returns the decision function of the sample for each class
             in the model.
             If decision_function_shape='ovr', the shape is (n_samples,
-            n_classes)
+            n_classes).
+
+        Notes
+        ------
+        If decision_function_shape='ovo', the function values are proportional
+        to the distance of the samples X to the separating hyperplane. If the
+        exact distances are required, divide the function values by the norm of
+        the weight vector (``coef_``). See also `this question
+        <https://stats.stackexchange.com/questions/14876/
+        interpreting-distance-from-hyperplane-in-svm>`_ for further details.
         """
         dec = self._decision_function(X)
         if self.decision_function_shape == 'ovr' and len(self.classes_) > 2:

From ea36f6ed8c6d39a168665e8320af35282fb7b2f9 Mon Sep 17 00:00:00 2001
From: Albert Thomas <albertthomas88@gmail.com>
Date: Tue, 11 Dec 2018 21:02:26 +0100
Subject: [PATCH 26/40] DOC change "cumulative density" to "cumulative
 distribution" (#12754)

---
 sklearn/preprocessing/data.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/sklearn/preprocessing/data.py b/sklearn/preprocessing/data.py
index c090852f7e02f..1560673b2bce4 100644
--- a/sklearn/preprocessing/data.py
+++ b/sklearn/preprocessing/data.py
@@ -1988,7 +1988,7 @@ class QuantileTransformer(BaseEstimator, TransformerMixin):
     (marginal) outliers: this is therefore a robust preprocessing scheme.
 
     The transformation is applied on each feature independently.
-    The cumulative density function of a feature is used to project the
+    The cumulative distribution function of a feature is used to project the
     original values. Features values of new/unseen data that fall below
     or above the fitted range will be mapped to the bounds of the output
     distribution. Note that this transform is non-linear. It may distort linear
@@ -2001,7 +2001,7 @@ class QuantileTransformer(BaseEstimator, TransformerMixin):
     ----------
     n_quantiles : int, optional (default=1000)
         Number of quantiles to be computed. It corresponds to the number
-        of landmarks used to discretize the cumulative density function.
+        of landmarks used to discretize the cumulative distribution function.
 
     output_distribution : str, optional (default='uniform')
         Marginal distribution for the transformed data. The choices are
@@ -2378,7 +2378,7 @@ def quantile_transform(X, axis=0, n_quantiles=1000,
     (marginal) outliers: this is therefore a robust preprocessing scheme.
 
     The transformation is applied on each feature independently.
-    The cumulative density function of a feature is used to project the
+    The cumulative distribution function of a feature is used to project the
     original values. Features values of new/unseen data that fall below
     or above the fitted range will be mapped to the bounds of the output
     distribution. Note that this transform is non-linear. It may distort linear
@@ -2398,7 +2398,7 @@ def quantile_transform(X, axis=0, n_quantiles=1000,
 
     n_quantiles : int, optional (default=1000)
         Number of quantiles to be computed. It corresponds to the number
-        of landmarks used to discretize the cumulative density function.
+        of landmarks used to discretize the cumulative distribution function.
 
     output_distribution : str, optional (default='uniform')
         Marginal distribution for the transformed data. The choices are

From 908898bdea766b09bb81ec1a993389a2b2166195 Mon Sep 17 00:00:00 2001
From: Nicolas Hug <contact@nicolas-hug.com>
Date: Wed, 12 Dec 2018 20:37:20 -0500
Subject: [PATCH 27/40] DOC Added roadmap.rst (#12761)

---
 doc/documentation.rst               |   5 +
 doc/roadmap.rst                     | 273 ++++++++++++++++++++++++++++
 doc/themes/scikit-learn/layout.html |   1 +
 3 files changed, 279 insertions(+)
 create mode 100644 doc/roadmap.rst

diff --git a/doc/documentation.rst b/doc/documentation.rst
index 8a972de4c4f93..05469edbe0b35 100644
--- a/doc/documentation.rst
+++ b/doc/documentation.rst
@@ -101,5 +101,10 @@ Documentation of scikit-learn |version|
                     scope or not well established enough for scikit-learn.
                     </blockquote>
                 </div>
+                <div class="span4 box">
+                    <h2><a href="roadmap.html">Roadmap</a></h2>
+                    <blockquote>Roadmap of the project.
+                    </blockquote>
+                </div>
 
             </div>
diff --git a/doc/roadmap.rst b/doc/roadmap.rst
new file mode 100644
index 0000000000000..a8334604395a2
--- /dev/null
+++ b/doc/roadmap.rst
@@ -0,0 +1,273 @@
+.. _roadmap:
+
+Roadmap
+=======
+
+Purpose of this document
+------------------------
+This document list general directions that core contributors are interested
+to see developed in scikit-learn. The fact that an item is listed here is in
+no way a promise that it will happen, as resources are limited. Rather, it
+is an indication that help is welcomed on this topic.
+
+Statement of purpose: Scikit-learn in 2018
+------------------------------------------
+Eleven years after the inception of Scikit-learn, much has changed in the
+world of machine learning. Key changes include:
+
+* Computational tools: The exploitation of GPUs, distributed programming
+  frameworks like Scala/Spark, etc.
+* High-level Python libraries for experimentation, processing and data
+  management: Jupyter notebook, Cython, Pandas, Dask, Numba...
+* Changes in the focus of machine learning research: artificial intelligence
+  applications (where input structure is key) with deep learning,
+  representation learning, reinforcement learning, domain transfer, etc.
+
+A more subtle change over the last decade is that, due to changing interests
+in ML, PhD students in machine learning are more likely to contribute to
+PyTorch, Dask, etc. than to Scikit-learn, so our contributor pool is very
+different to a decade ago.
+
+Scikit-learn remains very popular in practice for trying out canonical
+machine learning techniques, particularly for applications in experimental
+science and in data science. A lot of what we provide is now very mature.
+But it can be costly to maintain, and we cannot therefore include arbitrary
+new implementations. Yet Scikit-learn is also essential in defining an API
+framework for the development of interoperable machine learning components
+external to the core library.
+
+**Thus our main goals in this era are to**:
+
+* continue maintaining a high-quality, well-documented collection of canonical
+  tools for data processing and machine learning within the current scope
+  (i.e. rectangular data largely invariant to column and row order;
+  predicting targets with simple structure)
+* improve the ease for users to develop and publish external components
+* improve inter-operability with modern data science tools (e.g. Pandas, Dask)
+  and infrastructures (e.g. distributed processing)
+
+Many of the more fine-grained goals can be found under the `API tag
+<https://github.com/scikit-learn/scikit-learn/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-desc+label%3AAPI>`_
+on the issue tracker.
+
+Architectural / general goals
+-----------------------------
+The list is numbered not as an indication of the order of priority, but to
+make referring to specific points easier. Please add new entries only at the
+bottom.
+
+#. Everything in Scikit-learn should conform to our API contract
+
+   * `Pipeline <pipeline.Pipeline>` and `FeatureUnion` modify their input
+     parameters in fit. Fixing this requires making sure we have a good
+     grasp of their use cases to make sure all current functionality is
+     maintained. :issue:`8157` :issue:`7382`
+
+#. Improved handling of Pandas DataFrames and SparseDataFrames
+
+   * document current handling
+   * column reordering issue :issue:`7242`
+   * avoiding unnecessary conversion to ndarray
+   * returning DataFrames from transformers :issue:`5523`
+   * getting DataFrames from dataset loaders
+   * Sparse currently not considered
+
+#. Improved handling of categorical features
+
+   * Tree-based models should be able to handle both continuous and categorical
+     features :issue:`4899`
+   * In dataset loaders
+   * As generic transformers to be used with ColumnTransforms (e.g. ordinal
+     encoding supervised by correlation with target variable)
+
+#. Improved handling of missing data
+
+   * Making sure meta-estimators are lenient towards missing data
+   * Non-trivial imputers
+   * Learners directly handling missing data
+   * An amputation sample generator to make parts of a dataset go missing
+   * Handling mixtures of categorical and continuous variables
+
+#. Passing around information that is not (X, y): Sample properties
+
+   * We need to be able to pass sample weights to scorers in cross validation.
+   * We should have standard/generalised ways of passing sample-wise properties
+     around in meta-estimators. :issue:`4497` :issue:`7646`
+
+#. Passing around information that is not (X, y): Feature properties
+
+   * Feature names or descriptions should ideally be available to fit for, e.g.
+     . :issue:`6425` :issue:`6424`
+   * Per-feature handling (e.g. "is this a nominal / ordinal / English language
+     text?") should also not need to be provided to estimator constructors,
+     ideally, but should be available as metadata alongside X. :issue:`8480`
+
+#. Passing around information that is not (X, y): Target information
+
+   * We have problems getting the full set of classes to all components when
+     the data is split/sampled. :issue:`6231` :issue:`8100`
+   * We have no way to handle a mixture of categorical and continuous targets.
+
+#. Make it easier for external users to write Scikit-learn-compatible
+   components
+
+   * More flexible estimator checks that do not select by estimator name
+     :issue:`6599` :issue:`6715`
+   * Example of how to develop a meta-estimator
+   * More self-sufficient running of scikit-learn-contrib or a similar resource
+
+#. Support resampling and sample reduction
+
+   * Allow subsampling of majority classes (in a pipeline?) :issue:`3855`
+   * Implement random forests with resampling :issue:`8732`
+
+#. Better interfaces for interactive development
+
+   * __repr__ and HTML visualisations of estimators :issue:`6323`
+   * Include plotting tools, not just as examples. :issue:`9173`
+
+#. Improved tools for model diagnostics and basic inference
+
+   * partial dependence plots :issue:`5653`
+   * alternative feature importances implementations (e.g. methods or wrappers)
+   * better ways to handle validation sets when fitting
+   * better ways to find thresholds / create decision rules :issue:`8614`
+
+#. Better tools for selecting hyperparameters with transductive estimators
+
+   * Grid search and cross validation are not applicable to most clustering
+     tasks. Stability-based selection is more relevant.
+
+#. Improved tracking of fitting
+
+   * Verbose is not very friendly and should use a standard logging library
+     :issue:`6929`
+   * Callbacks or a similar system would facilitate logging and early stopping
+
+#. Use scipy BLAS Cython bindings
+
+   * This will make it possible to get rid of our partial copy of suboptimal
+     Atlas C-routines. :issue:`11638`
+   * This should speed up the Windows and Linux wheels
+
+#. Allow fine-grained parallelism in cython
+
+   * Now that we do not use fork-based multiprocessing in joblib anymore it's
+     possible to use the prange / openmp thread management which makes it
+     possible to have very efficient thread-based parallelism at the Cython
+     level. Example with K-Means: :issue:`11950`
+
+#. Distributed parallelism
+
+   * Joblib can now plug onto several backends, some of them can distribute the
+     computation across computers
+   * However, we want to stay high level in scikit-learn
+
+#. A way forward for more out of core
+
+   * Dask enables easy out-of-core computation. While the dask model probably
+     cannot be adaptable to all machine-learning algorithms, most machine
+     learning is on smaller data than ETL, hence we can maybe adapt to very
+     large scale while supporting only a fraction of the patterns.
+
+#. Better support for manual and automatic pipeline building
+
+   * Easier way to construct complex pipelines and valid search spaces
+     :issue:`7608` :issue:`5082` :issue:`8243`
+   * provide search ranges for common estimators??
+   * cf. `searchgrid <https://searchgrid.readthedocs.io/en/latest/>`_
+
+#. Support for working with pre-trained models
+
+   * Estimator "freezing". In particular, right now it's impossible to clone a
+     `CalibratedClassifierCV` with prefit. :issue:`8370`. :issue:`6451`
+
+#. Backwards-compatible de/serialization of some estimators
+
+   * Currently serialization (with pickle) breaks across versions. While we may
+     not be able to get around other limitations of pickle re security etc, it
+     would be great to offer cross-version safety from version 1.0. Note: Gael
+     and Olivier think that this can cause heavy maintenance burden and we
+     should manage the trade-offs. A possible alternative is presented in the
+     following point.
+
+#. Documentation and tooling for model lifecycle management
+
+   * Document good practices for model deployments and lifecycle: before
+     deploying a model: snapshot the code versions (numpy, scipy, scikit-learn,
+     custom code repo), the training script and an alias on how to retrieve
+     historical training data + snapshot a copy of a small validation set +
+     snapshot of the predictions (predicted probabilities for classifiers)
+     on that validation set.
+   * Document and tools to make it easy to manage upgrade of scikit-learn
+     versions:
+
+     * Try to load the old pickle, if it works, use the validation set
+       prediction snapshot to detect that the serialized model still behave
+       the same;
+     * If joblib.load / pickle.load not work, use the versioned control
+       training script + historical training set to retrain the model and use
+       the validation set prediction snapshot to assert that it is possible to
+       recover the previous predictive performance: if this is not the case
+       there is probably a bug in scikit-learn that needs to be reported.
+
+#. (Optional) Improve scikit-learn common tests suite to make sure that (at
+   least for frequently used) models have stable predictions across-versions
+   (to be discussed);
+
+   * Extend documentation to mention how to deploy models in Python-free
+     environments for instance  `ONNX <https://github.com/onnx/onnxmltools>`_.
+     and use the above best practices to assess predictive consistency between
+     scikit-learn and ONNX prediction functions on validation set.
+   * Document good practices to detect temporal distribution drift for deployed
+     model and good practices for re-training on fresh data without causing
+     catastrophic predictive performance regressions.
+
+#. More didactic documentation
+
+   * More and more options have been added to scikit-learn. As a result, the
+     documentation is crowded which makes it hard for beginners to get the big
+     picture. Some work could be done in prioritizing the information.
+
+Subpackage-specific goals
+-------------------------
+
+:mod:`sklearn.cluster`
+
+* kmeans variants for non-Euclidean distances, if we can show these have
+  benefits beyond hierarchical clustering.
+
+:mod:`sklearn.ensemble`
+
+* a stacking implementation
+* a binned feature histogram based and thread parallel implementation of
+  decision trees to compete with the performance of state of the art gradient
+  boosting like LightGBM.
+
+:mod:`sklearn.model_selection`
+
+* multi-metric scoring is slow :issue:`9326`
+* perhaps we want to be able to get back more than multiple metrics
+* the handling of random states in CV splitters is a poor design and
+  contradicts the validation of similar parameters in estimators.
+* exploit warm-starting and path algorithms so the benefits of `EstimatorCV`
+  objects can be accessed via `GridSearchCV` and used in Pipelines.
+  :issue:`1626`
+* Cross-validation should be able to be replaced by OOB estimates whenever a
+  cross-validation iterator is used.
+* Redundant computations in pipelines should be avoided (related to point
+  above) cf `daskml
+  <https://dask-ml.readthedocs.io/en/latest/hyper-parameter-search.html#avoid-repeated-work>`_
+
+:mod:`sklearn.neighbors`
+
+* Ability to substitute a custom/approximate/precomputed nearest neighbors
+  implementation for ours in all/most contexts that nearest neighbors are used
+  for learning. :issue:`10463`
+
+:mod:`sklearn.pipeline`
+
+* Performance issues with `Pipeline.memory`
+* see "Everything in Scikit-learn should conform to our API contract" above
+* Add a verbose option :issue:`10435`
+
diff --git a/doc/themes/scikit-learn/layout.html b/doc/themes/scikit-learn/layout.html
index 2add41c63b132..91c9b7336e095 100644
--- a/doc/themes/scikit-learn/layout.html
+++ b/doc/themes/scikit-learn/layout.html
@@ -90,6 +90,7 @@
             <li><a href="{{ pathto('glossary') }}">Glossary</a></li>
             <li><a href="{{ pathto('faq') }}">FAQ</a></li>
             <li><a href="{{ pathto('developers/contributing') }}">Contributing</a></li>
+            <li><a href="{{ pathto('roadmap') }}">Roadmap</a></li>
             <li class="divider"></li>
                 <script>if (VERSION_SUBDIR != "stable") document.write('<li><a href="http://scikit-learn.org/stable/documentation.html">Stable version</a></li>')</script>
                 <script>if (VERSION_SUBDIR != "dev") document.write('<li><a href="http://scikit-learn.org/dev/documentation.html">Development version</a></li>')</script>

From d3664b6e12e964a3a90f10d824fbf4f8af4806c2 Mon Sep 17 00:00:00 2001
From: Muhammad Hassaan Rafique <hassaan_rafique@hotmail.com>
Date: Thu, 13 Dec 2018 15:07:20 +0400
Subject: [PATCH 28/40] Update linear_model.rst (#12735)

* Update for Logistic-Regression Table Summary for L-BFGS

* Update for linear_model.rst Logistic-Regression L-BFGS summary.

* Update for Logistic-Regression Doc.

* Logistic-Regression Update

* Update linear_model.rst

* Update jnothman comments

* Update jnothman comments

* Update for jnothman comments.

* test

* test2

* test3

* test4

* Update for Comments from amueller and jnothman

* Update for Comments from amueller and jnothman
---
 doc/modules/linear_model.rst | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/doc/modules/linear_model.rst b/doc/modules/linear_model.rst
index ae2ce9ee1a8ff..4225863ac9384 100644
--- a/doc/modules/linear_model.rst
+++ b/doc/modules/linear_model.rst
@@ -775,7 +775,12 @@ The "saga" solver [7]_ is a variant of "sag" that also supports the
 non-smooth `penalty="l1"` option. This is therefore the solver of choice
 for sparse multinomial logistic regression.
 
-In a nutshell, the following table summarizes the penalties supported by each solver:
+The "lbfgs" is an optimization algorithm that approximates the 
+Broyden–Fletcher–Goldfarb–Shanno algorithm [8]_, which belongs to
+quasi-Newton methods. The "lbfgs" solver is recommended for use for
+small data-sets but for larger datasets its performance suffers. [9]_
+
+The following table summarizes the penalties supported by each solver:
 
 +------------------------------+-----------------+-------------+-----------------+-----------+------------+
 |                              |                       **Solvers**                                        |
@@ -799,8 +804,8 @@ In a nutshell, the following table summarizes the penalties supported by each so
 | Robust to unscaled datasets  |       yes       |     yes     |       yes       |    no     |    no      |
 +------------------------------+-----------------+-------------+-----------------+-----------+------------+
 
-The "saga" solver is often the best choice but requires scaling. The "liblinear" solver is
-used by default for historical reasons.
+The "saga" solver is often the best choice but requires scaling. The
+"lbfgs" solver is used by default for historical reasons.
 
 For large dataset, you may also consider using :class:`SGDClassifier`
 with 'log' loss.
@@ -855,6 +860,12 @@ loss.
 
     .. [7] Aaron Defazio, Francis Bach, Simon Lacoste-Julien: `SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives. <https://arxiv.org/abs/1407.0202>`_
 
+    .. [8] https://en.wikipedia.org/wiki/Broyden%E2%80%93Fletcher%E2%80%93Goldfarb%E2%80%93Shanno_algorithm
+
+    .. [9] `"Performance Evaluation of Lbfgs vs other solvers"
+            <http://www.fuzihao.org/blog/2016/01/16/Comparison-of-Gradient-Descent-Stochastic-Gradient-Descent-and-L-BFGS/>`_
+
+
 Stochastic Gradient Descent - SGD
 =================================
 

From 44a0c801f17c62c2d73c91baf095041a2c1487cd Mon Sep 17 00:00:00 2001
From: Andreas Mueller <t3kcit@gmail.com>
Date: Thu, 13 Dec 2018 21:30:58 -0500
Subject: [PATCH 29/40] DOC Explain solver choices for LogisticRegression
 (#12768)

---
 doc/modules/linear_model.rst | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/doc/modules/linear_model.rst b/doc/modules/linear_model.rst
index 4225863ac9384..d0a5ad8401914 100644
--- a/doc/modules/linear_model.rst
+++ b/doc/modules/linear_model.rst
@@ -804,11 +804,10 @@ The following table summarizes the penalties supported by each solver:
 | Robust to unscaled datasets  |       yes       |     yes     |       yes       |    no     |    no      |
 +------------------------------+-----------------+-------------+-----------------+-----------+------------+
 
-The "saga" solver is often the best choice but requires scaling. The
-"lbfgs" solver is used by default for historical reasons.
-
+The "lbfgs" solver is used by default for its robustness. For large datasets
+the "saga" solver is usually faster.
 For large dataset, you may also consider using :class:`SGDClassifier`
-with 'log' loss.
+with 'log' loss, which might be even faster but require more tuning.
 
 .. topic:: Examples:
 

From c51fd04839d654e2c4b2265094c408be50c73411 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Fri, 14 Dec 2018 21:04:02 +0800
Subject: [PATCH 30/40] MNT flake8 in r2_score examples (#12773)

---
 sklearn/metrics/regression.py | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/sklearn/metrics/regression.py b/sklearn/metrics/regression.py
index e9084a4276e18..5bb40c5cf59dc 100644
--- a/sklearn/metrics/regression.py
+++ b/sklearn/metrics/regression.py
@@ -514,19 +514,19 @@ def r2_score(y_true, y_pred, sample_weight=None,
     0.948...
     >>> y_true = [[0.5, 1], [-1, 1], [7, -6]]
     >>> y_pred = [[0, 2], [-1, 2], [8, -5]]
-    >>> r2_score(y_true, y_pred, multioutput='variance_weighted')
-    ... # doctest: +ELLIPSIS
+    >>> r2_score(y_true, y_pred,
+    ...          multioutput='variance_weighted') # doctest: +ELLIPSIS
     0.938...
-    >>> y_true = [1,2,3]
-    >>> y_pred = [1,2,3]
+    >>> y_true = [1, 2, 3]
+    >>> y_pred = [1, 2, 3]
     >>> r2_score(y_true, y_pred)
     1.0
-    >>> y_true = [1,2,3]
-    >>> y_pred = [2,2,2]
+    >>> y_true = [1, 2, 3]
+    >>> y_pred = [2, 2, 2]
     >>> r2_score(y_true, y_pred)
     0.0
-    >>> y_true = [1,2,3]
-    >>> y_pred = [3,2,1]
+    >>> y_true = [1, 2, 3]
+    >>> y_pred = [3, 2, 1]
     >>> r2_score(y_true, y_pred)
     -3.0
     """

From 9ad98479cffcb16bec1e0acc3e3cffe9bd485781 Mon Sep 17 00:00:00 2001
From: Feda Curic <feda.curic@gmail.com>
Date: Fri, 14 Dec 2018 16:40:49 +0100
Subject: [PATCH 31/40] EXA Avoiding TypeError when using --all_categories in
 plot_document_classification_20newsgroups.py (#12770)

---
 examples/text/plot_document_classification_20newsgroups.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/examples/text/plot_document_classification_20newsgroups.py b/examples/text/plot_document_classification_20newsgroups.py
index 3884393818157..b761c8241f300 100644
--- a/examples/text/plot_document_classification_20newsgroups.py
+++ b/examples/text/plot_document_classification_20newsgroups.py
@@ -145,7 +145,7 @@ def size_mb(docs):
     len(data_train.data), data_train_size_mb))
 print("%d documents - %0.3fMB (test set)" % (
     len(data_test.data), data_test_size_mb))
-print("%d categories" % len(categories))
+print("%d categories" % len(target_names))
 print()
 
 # split a training set and a test set

From cc846b45e3eb29bb8bdf9439000b7f91b9c37692 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Sat, 15 Dec 2018 09:59:40 +0800
Subject: [PATCH 32/40] DOC Remove what's new entry for unreleased feature

---
 doc/whats_new/v0.20.rst | 1 -
 1 file changed, 1 deletion(-)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 4581d500ab307..97e6caee3b577 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -32,7 +32,6 @@ Changelog
   unexpected error when columns is pandas Index or pandas Series.
   :issue:`12704` by :user:`Hanmin Qin <qinhanmin2014>`.
 
-
 :mod:`sklearn.neighbors`
 ........................
 

From da99f5c41a10097ad11c91b659198751d440d953 Mon Sep 17 00:00:00 2001
From: Katrin Leinweber <9948149+katrinleinweber@users.noreply.github.com>
Date: Sat, 15 Dec 2018 16:36:19 +0100
Subject: [PATCH 33/40] DOC Hyperlink DOIs to preferred resolver (#12792)

---
 doc/modules/clustering.rst | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/doc/modules/clustering.rst b/doc/modules/clustering.rst
index bf5fabe8386cb..12e48d24404d1 100644
--- a/doc/modules/clustering.rst
+++ b/doc/modules/clustering.rst
@@ -1680,12 +1680,12 @@ Drawbacks
    "A Cluster Separation Measure"
    IEEE Transactions on Pattern Analysis and Machine Intelligence.
    PAMI-1 (2): 224-227.
-   `doi:10.1109/TPAMI.1979.4766909 <http://dx.doi.org/10.1109/TPAMI.1979.4766909>`_.
+   `doi:10.1109/TPAMI.1979.4766909 <https://doi.org/10.1109/TPAMI.1979.4766909>`_.
 
  * Halkidi, Maria; Batistakis, Yannis; Vazirgiannis, Michalis (2001).
    "On Clustering Validation Techniques"
    Journal of Intelligent Information Systems, 17(2-3), 107-145.
-   `doi:10.1023/A:1012801612483 <http://dx.doi.org/10.1023/A:1012801612483>`_.
+   `doi:10.1023/A:1012801612483 <https://doi.org/10.1023/A:1012801612483>`_.
 
  * `Wikipedia entry for Davies-Bouldin index
    <https://en.wikipedia.org/wiki/Davies–Bouldin_index>`_.

From bb30e4ce5d1574543f80feddbb878ea43313f9b7 Mon Sep 17 00:00:00 2001
From: adanhawth <36526095+adanhawth@users.noreply.github.com>
Date: Sat, 15 Dec 2018 20:07:03 -0500
Subject: [PATCH 34/40] DOC Fix LaTeX text{} block escaping of underscores in
 doc-modules-model_evaluation.rst (#12788)

---
 doc/modules/model_evaluation.rst | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/doc/modules/model_evaluation.rst b/doc/modules/model_evaluation.rst
index 07c719c87cef9..3e54e18c22456 100644
--- a/doc/modules/model_evaluation.rst
+++ b/doc/modules/model_evaluation.rst
@@ -440,10 +440,10 @@ the total number of predictions).
 
 In contrast, if the conventional accuracy is above chance only because the
 classifier takes advantage of an imbalanced test set, then the balanced
-accuracy, as appropriate, will drop to :math:`\frac{1}{\text{n\_classes}}`.
+accuracy, as appropriate, will drop to :math:`\frac{1}{n\_classes}`.
 
 The score ranges from 0 to 1, or when ``adjusted=True`` is used, it rescaled to
-the range :math:`\frac{1}{1 - \text{n\_classes}}` to 1, inclusive, with
+the range :math:`\frac{1}{1 - n\_classes}` to 1, inclusive, with
 performance at random scoring 0.
 
 If :math:`y_i` is the true value of the :math:`i`-th sample, and :math:`w_i`
@@ -463,7 +463,7 @@ defined as:
 
 With ``adjusted=True``, balanced accuracy reports the relative increase from
 :math:`\texttt{balanced-accuracy}(y, \mathbf{0}, w) =
-\frac{1}{\text{n\_classes}}`.  In the binary case, this is also known as
+\frac{1}{n\_classes}`.  In the binary case, this is also known as
 `*Youden's J statistic* <https://en.wikipedia.org/wiki/Youden%27s_J_statistic>`_,
 or *informedness*.
 
@@ -1433,7 +1433,7 @@ score associated with each label
 the ranking loss is defined as
 
 .. math::
-  \text{ranking\_loss}(y, \hat{f}) =  \frac{1}{n_{\text{samples}}}
+  ranking\_loss(y, \hat{f}) =  \frac{1}{n_{\text{samples}}}
     \sum_{i=0}^{n_{\text{samples}} - 1} \frac{1}{||y_i||_0(n_\text{labels} - ||y_i||_0)}
     \left|\left\{(k, l): \hat{f}_{ik} \leq \hat{f}_{il}, y_{ik} = 1, y_{il} = 0 \right\}\right|
 
@@ -1509,7 +1509,7 @@ then the explained variance is estimated as follow:
 
 .. math::
 
-  \texttt{explained\_{}variance}(y, \hat{y}) = 1 - \frac{Var\{ y - \hat{y}\}}{Var\{y\}}
+  explained\_{}variance(y, \hat{y}) = 1 - \frac{Var\{ y - \hat{y}\}}{Var\{y\}}
 
 The best possible score is 1.0, lower values are worse.
 

From 895aeb58ee90bd3eb8c32802bff6b6acfa072ef6 Mon Sep 17 00:00:00 2001
From: Reshama Shaikh <rs2715@stern.nyu.edu>
Date: Mon, 17 Dec 2018 02:15:24 -0500
Subject: [PATCH 35/40] DOC Added description for classes_ in
 LogisticRegression/LogisticRegressionCV documentation (#12795)

---
 sklearn/linear_model/logistic.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/sklearn/linear_model/logistic.py b/sklearn/linear_model/logistic.py
index 2f24c76397ccd..5870a7672ce35 100644
--- a/sklearn/linear_model/logistic.py
+++ b/sklearn/linear_model/logistic.py
@@ -1142,6 +1142,9 @@ class LogisticRegression(BaseEstimator, LinearClassifierMixin,
     Attributes
     ----------
 
+    classes_ : array, shape (n_classes, )
+        A list of class labels known to the classifier.
+
     coef_ : array, shape (1, n_features) or (n_classes, n_features)
         Coefficient of the features in the decision function.
 
@@ -1594,6 +1597,9 @@ class LogisticRegressionCV(LogisticRegression, BaseEstimator,
 
     Attributes
     ----------
+    classes_ : array, shape (n_classes, )
+        A list of class labels known to the classifier.
+
     coef_ : array, shape (1, n_features) or (n_classes, n_features)
         Coefficient of the features in the decision function.
 

From 713c1643f08a986a238aef4de679e7ba0c4e61b1 Mon Sep 17 00:00:00 2001
From: jeremiedbb <34657725+jeremiedbb@users.noreply.github.com>
Date: Mon, 17 Dec 2018 20:58:40 +0100
Subject: [PATCH 36/40] FIX pairwise distances with 'seuclidean' or
 'mahalanobis' metrics (#12701)

Fixes #12672
---
 doc/whats_new/v0.20.rst                | 10 +++++++
 sklearn/metrics/pairwise.py            | 26 ++++++++++++++++++
 sklearn/metrics/tests/test_pairwise.py | 38 ++++++++++++++++++++++++++
 3 files changed, 74 insertions(+)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 97e6caee3b577..6329d9b232149 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -46,6 +46,16 @@ Changelog
   data, which raised an error in 0.20.0, now returns the expected output again.
   :issue:`12699` by `Joris Van den Bossche`_.
 
+:mod:`sklearn.metrics`
+......................
+
+- |Fix| Fixed a bug in :func:`metrics.pairwise_distances` and
+  :func:`metrics.pairwise_distances_chunked` where parameters ``V`` of
+  ``"seuclidean"`` and ``VI`` of ``"mahalanobis"`` metrics were computed after
+  the data was split into chunks instead of being pre-computed on whole data.
+  :issue:`12701` by :user:`Jeremie du Boisberranger <jeremiedbb>`.
+
+
 .. _changes_0_20_1:
 
 Version 0.20.1
diff --git a/sklearn/metrics/pairwise.py b/sklearn/metrics/pairwise.py
index 6258881cde2c1..19a5a25fff75c 100644
--- a/sklearn/metrics/pairwise.py
+++ b/sklearn/metrics/pairwise.py
@@ -1136,6 +1136,24 @@ def _check_chunk_size(reduced, chunk_size):
                           chunk_size))
 
 
+def _precompute_metric_params(X, Y, metric=None, **kwds):
+    """Precompute data-derived metric parameters if not provided
+    """
+    if metric == "seuclidean" and 'V' not in kwds:
+        if X is Y:
+            V = np.var(X, axis=0, ddof=1)
+        else:
+            V = np.var(np.vstack([X, Y]), axis=0, ddof=1)
+        return {'V': V}
+    if metric == "mahalanobis" and 'VI' not in kwds:
+        if X is Y:
+            VI = np.linalg.inv(np.cov(X.T)).T
+        else:
+            VI = np.linalg.inv(np.cov(np.vstack([X, Y]).T)).T
+        return {'VI': VI}
+    return {}
+
+
 def pairwise_distances_chunked(X, Y=None, reduce_func=None,
                                metric='euclidean', n_jobs=None,
                                working_memory=None, **kwds):
@@ -1271,6 +1289,10 @@ def pairwise_distances_chunked(X, Y=None, reduce_func=None,
                                         working_memory=working_memory)
         slices = gen_batches(n_samples_X, chunk_n_rows)
 
+    # precompute data-derived metric params
+    params = _precompute_metric_params(X, Y, metric=metric, **kwds)
+    kwds.update(**params)
+
     for sl in slices:
         if sl.start == 0 and sl.stop == n_samples_X:
             X_chunk = X  # enable optimised paths for X is Y
@@ -1398,6 +1420,10 @@ def pairwise_distances(X, Y=None, metric="euclidean", n_jobs=None, **kwds):
         dtype = bool if metric in PAIRWISE_BOOLEAN_FUNCTIONS else None
         X, Y = check_pairwise_arrays(X, Y, dtype=dtype)
 
+        # precompute data-derived metric params
+        params = _precompute_metric_params(X, Y, metric=metric, **kwds)
+        kwds.update(**params)
+
         if effective_n_jobs(n_jobs) == 1 and X is Y:
             return distance.squareform(distance.pdist(X, metric=metric,
                                                       **kwds))
diff --git a/sklearn/metrics/tests/test_pairwise.py b/sklearn/metrics/tests/test_pairwise.py
index d19f3fc873500..80266f1318b89 100644
--- a/sklearn/metrics/tests/test_pairwise.py
+++ b/sklearn/metrics/tests/test_pairwise.py
@@ -5,9 +5,12 @@
 
 from scipy.sparse import dok_matrix, csr_matrix, issparse
 from scipy.spatial.distance import cosine, cityblock, minkowski, wminkowski
+from scipy.spatial.distance import cdist, pdist, squareform
 
 import pytest
 
+from sklearn import config_context
+
 from sklearn.utils.testing import assert_greater
 from sklearn.utils.testing import assert_array_almost_equal
 from sklearn.utils.testing import assert_allclose
@@ -893,3 +896,38 @@ def test_check_preserve_type():
                                                    XB.astype(np.float))
     assert_equal(XA_checked.dtype, np.float)
     assert_equal(XB_checked.dtype, np.float)
+
+
+@pytest.mark.parametrize("n_jobs", [1, 2])
+@pytest.mark.parametrize("metric", ["seuclidean", "mahalanobis"])
+@pytest.mark.parametrize("dist_function",
+                         [pairwise_distances, pairwise_distances_chunked])
+@pytest.mark.parametrize("y_is_x", [True, False], ids=["Y is X", "Y is not X"])
+def test_pairwise_distances_data_derived_params(n_jobs, metric, dist_function,
+                                                y_is_x):
+    # check that pairwise_distances give the same result in sequential and
+    # parallel, when metric has data-derived parameters.
+    with config_context(working_memory=0.1):  # to have more than 1 chunk
+        rng = np.random.RandomState(0)
+        X = rng.random_sample((1000, 10))
+
+        if y_is_x:
+            Y = X
+            expected_dist_default_params = squareform(pdist(X, metric=metric))
+            if metric == "seuclidean":
+                params = {'V': np.var(X, axis=0, ddof=1)}
+            else:
+                params = {'VI': np.linalg.inv(np.cov(X.T)).T}
+        else:
+            Y = rng.random_sample((1000, 10))
+            expected_dist_default_params = cdist(X, Y, metric=metric)
+            if metric == "seuclidean":
+                params = {'V': np.var(np.vstack([X, Y]), axis=0, ddof=1)}
+            else:
+                params = {'VI': np.linalg.inv(np.cov(np.vstack([X, Y]).T)).T}
+
+        expected_dist_explicit_params = cdist(X, Y, metric=metric, **params)
+        dist = np.vstack(dist_function(X, Y, metric=metric, n_jobs=n_jobs))
+
+        assert_allclose(dist, expected_dist_explicit_params)
+        assert_allclose(dist, expected_dist_default_params)

From 5ab850a976f657c45f9b4e4282e3d77c4c868629 Mon Sep 17 00:00:00 2001
From: Andreas Mueller <andreas.mueller@columbia.edu>
Date: Mon, 17 Dec 2018 15:34:53 -0500
Subject: [PATCH 37/40] add contributors, update website

---
 doc/index.rst           |  2 ++
 doc/whats_new/v0.20.rst | 13 ++++++++++++-
 2 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/doc/index.rst b/doc/index.rst
index c665890474686..16c984c38f251 100644
--- a/doc/index.rst
+++ b/doc/index.rst
@@ -209,6 +209,8 @@
                     </li>
                     <li><strong>Scikit-learn 0.21 will drop support for Python 2.7 and Python 3.4.</strong>
                     </li>
+                    <li><em>December 2018.</em> scikit-learn 0.20.2 is available for download (<a href="whats_new.html#version-0-20-2">Changelog</a>).
+                    </li>
                     <li><em>November 2018.</em> scikit-learn 0.20.1 is available for download (<a href="whats_new.html#version-0-20-1">Changelog</a>).
                     </li>
                     <li><em>September 2018.</em> scikit-learn 0.20.0 is available for download (<a href="whats_new.html#version-0-20-0">Changelog</a>).
diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 6329d9b232149..76337249c9cde 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -7,7 +7,7 @@
 Version 0.20.2
 ==============
 
-**XX, 2019**
+**December 20, 2018**
 
 This is a bug-fix release with some minor documentation improvements and
 enhancements to features released in 0.20.0.
@@ -55,6 +55,17 @@ Changelog
   the data was split into chunks instead of being pre-computed on whole data.
   :issue:`12701` by :user:`Jeremie du Boisberranger <jeremiedbb>`.
 
+Code and Documentation Contributors
+-----------------------------------
+
+With thanks to:
+
+
+adanhawth, Adrin Jalali, Albert Thomas, Andreas Mueller, Dan Stine, Feda Curic,
+Hanmin Qin, Jan S, jeremiedbb, Joel Nothman, Joris Van den Bossche,
+josephsalmon, Katrin Leinweber, Loic Esteve, Muhammad Hassaan Rafique, Nicolas
+Hug, Olivier Grisel, Paul Paczuski, Reshama Shaikh, Sam Waterbury, Shivam
+Kotwalia, Thomas Fan
 
 .. _changes_0_20_1:
 

From ad8764d6b889db44eabb377b3a11554dc70d780c Mon Sep 17 00:00:00 2001
From: Joel Nothman <joel.nothman@gmail.com>
Date: Wed, 19 Dec 2018 01:06:15 +1100
Subject: [PATCH 38/40] DOC reorder what's new 0.20.2

---
 doc/whats_new/v0.20.rst | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index 22ddf3296e27c..b11056b679f00 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -32,6 +32,15 @@ Changelog
   unexpected error when columns is pandas Index or pandas Series.
   :issue:`12704` by :user:`Hanmin Qin <qinhanmin2014>`.
 
+:mod:`sklearn.metrics`
+......................
+
+- |Fix| Fixed a bug in :func:`metrics.pairwise_distances` and
+  :func:`metrics.pairwise_distances_chunked` where parameters ``V`` of
+  ``"seuclidean"`` and ``VI`` of ``"mahalanobis"`` metrics were computed after
+  the data was split into chunks instead of being pre-computed on whole data.
+  :issue:`12701` by :user:`Jeremie du Boisberranger <jeremiedbb>`.
+
 :mod:`sklearn.neighbors`
 ........................
 
@@ -46,15 +55,6 @@ Changelog
   data, which raised an error in 0.20.0, now returns the expected output again.
   :issue:`12699` by `Joris Van den Bossche`_.
 
-:mod:`sklearn.metrics`
-......................
-
-- |Fix| Fixed a bug in :func:`metrics.pairwise_distances` and
-  :func:`metrics.pairwise_distances_chunked` where parameters ``V`` of
-  ``"seuclidean"`` and ``VI`` of ``"mahalanobis"`` metrics were computed after
-  the data was split into chunks instead of being pre-computed on whole data.
-  :issue:`12701` by :user:`Jeremie du Boisberranger <jeremiedbb>`.
-
 Code and Documentation Contributors
 -----------------------------------
 

From 8da0d2328e65bce83d8eb9170700737a7e2e265f Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Tue, 18 Dec 2018 22:41:38 +0800
Subject: [PATCH 39/40] TST np.vstack won't support generator in the future
 (#12816)

---
 sklearn/metrics/tests/test_pairwise.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/sklearn/metrics/tests/test_pairwise.py b/sklearn/metrics/tests/test_pairwise.py
index 80266f1318b89..aa7b0cc809bbc 100644
--- a/sklearn/metrics/tests/test_pairwise.py
+++ b/sklearn/metrics/tests/test_pairwise.py
@@ -927,7 +927,8 @@ def test_pairwise_distances_data_derived_params(n_jobs, metric, dist_function,
                 params = {'VI': np.linalg.inv(np.cov(np.vstack([X, Y]).T)).T}
 
         expected_dist_explicit_params = cdist(X, Y, metric=metric, **params)
-        dist = np.vstack(dist_function(X, Y, metric=metric, n_jobs=n_jobs))
+        dist = np.vstack(tuple(dist_function(X, Y,
+                                             metric=metric, n_jobs=n_jobs)))
 
         assert_allclose(dist, expected_dist_explicit_params)
         assert_allclose(dist, expected_dist_default_params)

From a941c2e25ebaa3060c71a2390aa525f79028b8a5 Mon Sep 17 00:00:00 2001
From: Hanmin Qin <qinhanmin2005@sina.com>
Date: Wed, 19 Dec 2018 06:18:38 +0800
Subject: [PATCH 40/40] DOC Update changed models for 0.20.2 (#12810)

---
 doc/whats_new/v0.20.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
index b11056b679f00..13782c0ccfb30 100644
--- a/doc/whats_new/v0.20.rst
+++ b/doc/whats_new/v0.20.rst
@@ -21,6 +21,7 @@ occurs due to changes in the modelling logic (bug fixes or enhancements), or in
 random sampling procedures.
 
 - :mod:`sklearn.neighbors` when ``metric=='jaccard'`` (bug fix)
+- use of ``'seuclidean'`` or ``'mahalanobis'`` metrics in some cases (bug fix)
 
 Changelog
 ---------