From 3d0236991cb9db3e64defe7b649c218c4f7cafee Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Mon, 8 Mar 2021 14:22:22 +0100
Subject: [PATCH 01/16] raise an error when the 'pred_decision' shape is not
 consistent with number of classes with a multiclass target case. New test in
 order to check this situation in 'test_classification.py'

---
 sklearn/metrics/_classification.py           |  8 ++++++++
 sklearn/metrics/tests/test_classification.py | 19 +++++++++++++++++++
 2 files changed, 27 insertions(+)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index 708bde662e765..f12e76b2b55f0 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2371,6 +2371,14 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     y_true = column_or_1d(y_true)
     y_true_unique = np.unique(labels if labels is not None else y_true)
     if y_true_unique.size > 2:
+        if (pred_decision.ndim == 1) or \
+                (labels is not None and pred_decision.ndim > 1 and
+                 np.size(y_true_unique) != pred_decision.shape[1]):
+            raise ValueError("The shape of pred_decision is not "
+                             "consistent with the number of classes. "
+                             "pred_decision shape must be "
+                             "(n_samples, n_classes) with "
+                             "multiclass target")
         if (labels is None and pred_decision.ndim > 1 and
                 (np.size(y_true_unique) != pred_decision.shape[1])):
             raise ValueError("Please include all labels in y_true "
diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index c32e9c89ada47..9e766d6e7d9ee 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -2135,6 +2135,25 @@ def test_hinge_loss_multiclass_missing_labels_with_labels_none():
         hinge_loss(y_true, pred_decision)
 
 
+def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
+    y_true = [2, 1, 0, 1, 0, 1, 1]
+    pred_decision = np.array([0, 1, 2, 1, 0, 2, 1])
+    error_message = ("The shape of pred_decision is not "
+                     "consistent with the number of classes. "
+                     "pred_decision shape must be "
+                     "(n_samples, n_classes) with "
+                     "multiclass target")
+
+    assert_raise_message(ValueError, error_message, hinge_loss,
+                         y_true=y_true, pred_decision=pred_decision)
+    pred_decision = [[0, 1], [0, 1], [0, 1], [0, 1],
+                     [2, 0], [0, 1], [1, 0]]
+    labels = [0, 1, 2]
+    assert_raise_message(ValueError, error_message, hinge_loss,
+                         y_true=y_true, pred_decision=pred_decision,
+                         labels=labels)
+
+
 def test_hinge_loss_multiclass_with_missing_labels():
     pred_decision = np.array([
         [+0.36, -0.17, -0.58, -0.99],

From 86197f34924f492576eecef12eca1a7e4357026b Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Mon, 8 Mar 2021 14:41:46 +0100
Subject: [PATCH 02/16] add the asset_raise_message import

---
 sklearn/metrics/tests/test_classification.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 9e766d6e7d9ee..55873f4f934a3 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -15,6 +15,7 @@
 from sklearn.datasets import make_multilabel_classification
 from sklearn.preprocessing import label_binarize, LabelBinarizer
 from sklearn.utils.validation import check_random_state
+from sklearn.utils._testing import assert_raise_message
 from sklearn.utils._testing import assert_almost_equal
 from sklearn.utils._testing import assert_array_equal
 from sklearn.utils._testing import assert_array_almost_equal

From feee8f5582c2b25e1cb3d846b5b42012834c127c Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Mon, 8 Mar 2021 15:22:02 +0100
Subject: [PATCH 03/16] add changelog

---
 doc/whats_new/v1.0.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/doc/whats_new/v1.0.rst b/doc/whats_new/v1.0.rst
index 3086d91b28f5d..575fbdbdd2813 100644
--- a/doc/whats_new/v1.0.rst
+++ b/doc/whats_new/v1.0.rst
@@ -134,6 +134,11 @@ Changelog
   class methods and will be removed in 1.2.
   :pr:`18543` by `Guillaume Lemaitre`_.
 
+- |Enhancement| A fix to raise an error when ``pred_decision`` is 1d
+  whereas it is a multiclass classification or when ``pred_decision``
+  parameter is not consistent with ``labels`` parameter.
+  :pr:`19643` by :user:`Pierre Attard <PierreAttard>`.
+
 :mod:`sklearn.naive_bayes`
 ..........................
 

From cc6462f79c91a0ecd1412441d2c1d8126a0b76a4 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Wed, 10 Mar 2021 20:41:48 +0100
Subject: [PATCH 04/16] cleaner test in order to raise the right error

---
 sklearn/metrics/_classification.py | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index f12e76b2b55f0..d772dab0a4853 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2371,18 +2371,18 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     y_true = column_or_1d(y_true)
     y_true_unique = np.unique(labels if labels is not None else y_true)
     if y_true_unique.size > 2:
-        if (pred_decision.ndim == 1) or \
-                (labels is not None and pred_decision.ndim > 1 and
-                 np.size(y_true_unique) != pred_decision.shape[1]):
-            raise ValueError("The shape of pred_decision is not "
-                             "consistent with the number of classes. "
-                             "pred_decision shape must be "
-                             "(n_samples, n_classes) with "
-                             "multiclass target")
-        if (labels is None and pred_decision.ndim > 1 and
-                (np.size(y_true_unique) != pred_decision.shape[1])):
-            raise ValueError("Please include all labels in y_true "
-                             "or pass labels as third argument")
+        if pred_decision.ndim <= 1 or \
+                y_true_unique.size != pred_decision.shape[1]:
+            if labels is not None or pred_decision.ndim <= 1:
+                raise ValueError("The shape of pred_decision is not "
+                                 "consistent with the number of classes. "
+                                 "pred_decision shape must be "
+                                 "(n_samples, n_classes) with "
+                                 "multiclass target")
+            else:
+                raise ValueError("Please include all labels in y_true "
+                                 "or pass labels as third argument")
+
         if labels is None:
             labels = y_true_unique
         le = LabelEncoder()

From 50b2878490bad7552101ebe7c1191e66a65ca81b Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 09:13:48 +0100
Subject: [PATCH 05/16] all tests for valid multiclass decision inside a
 function.

---
 sklearn/metrics/_classification.py | 56 ++++++++++++++++++++++++------
 1 file changed, 45 insertions(+), 11 deletions(-)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index d772dab0a4853..7fa3fd19ce616 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2285,6 +2285,48 @@ def log_loss(y_true, y_pred, *, eps=1e-15, normalize=True, sample_weight=None,
     return _weighted_sum(loss, sample_weight, normalize)
 
 
+def _check_valid_multiclass_decision_shape(y_true_unique, pred_decision,
+                                           labels):
+    """Check if the inputs are consistent.
+
+    If it is a multiclass problem (it means thare are more than 2 classes
+    to classify), `pred_decision` parameter can't be 1d array. The column
+    number must be equal to classes number.
+
+    If not every classes are present in `y_true`, `labels` parameter
+    must be filled in.
+
+    Parameters
+    ----------
+    y_true_unique : array-like
+
+    pred_decision : array-like
+
+    labels : array-like
+    """
+
+    if pred_decision.ndim <= 1:
+        raise ValueError("The shape of pred_decision can not be 1d array"
+                         "with a multiclass target. "
+                         "pred_decision shape must be (n_samples, n_classes)")
+
+    invalid_decision_shape = pred_decision.ndim > 1 and \
+                             y_true_unique.size != pred_decision.shape[1]
+
+    if labels is None:
+        if invalid_decision_shape:
+            raise ValueError("Please include all labels in y_true "
+                             "or pass labels as third argument")
+    elif invalid_decision_shape:
+        raise ValueError("The shape of pred_decision is not "
+                         "consistent with the number of classes. "
+                         "pred_decision shape must be "
+                         "(n_samples, n_classes) with "
+                         "multiclass target")
+
+    return
+
+
 @_deprecate_positional_args
 def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     """Average hinge loss (non-regularized).
@@ -2370,19 +2412,11 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     pred_decision = check_array(pred_decision, ensure_2d=False)
     y_true = column_or_1d(y_true)
     y_true_unique = np.unique(labels if labels is not None else y_true)
+
     if y_true_unique.size > 2:
-        if pred_decision.ndim <= 1 or \
-                y_true_unique.size != pred_decision.shape[1]:
-            if labels is not None or pred_decision.ndim <= 1:
-                raise ValueError("The shape of pred_decision is not "
-                                 "consistent with the number of classes. "
-                                 "pred_decision shape must be "
-                                 "(n_samples, n_classes) with "
-                                 "multiclass target")
-            else:
-                raise ValueError("Please include all labels in y_true "
-                                 "or pass labels as third argument")
 
+        _check_valid_multiclass_decision_shape(y_true_unique, pred_decision,
+                                               labels)
         if labels is None:
             labels = y_true_unique
         le = LabelEncoder()

From f8554005a3e3819d576e86066775d6e6efa4a8d9 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 09:15:26 +0100
Subject: [PATCH 06/16] `assert_raise_message` replaced by `pytest.raises`.
 import of assert_raise_message removed

---
 sklearn/metrics/tests/test_classification.py | 25 +++++++++++---------
 1 file changed, 14 insertions(+), 11 deletions(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 55873f4f934a3..28c9524960bd4 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -15,7 +15,6 @@
 from sklearn.datasets import make_multilabel_classification
 from sklearn.preprocessing import label_binarize, LabelBinarizer
 from sklearn.utils.validation import check_random_state
-from sklearn.utils._testing import assert_raise_message
 from sklearn.utils._testing import assert_almost_equal
 from sklearn.utils._testing import assert_array_equal
 from sklearn.utils._testing import assert_array_almost_equal
@@ -2137,22 +2136,26 @@ def test_hinge_loss_multiclass_missing_labels_with_labels_none():
 
 
 def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
+    # test for inconsistency between multiclass problem and pred_decision argument
     y_true = [2, 1, 0, 1, 0, 1, 1]
     pred_decision = np.array([0, 1, 2, 1, 0, 2, 1])
-    error_message = ("The shape of pred_decision is not "
-                     "consistent with the number of classes. "
-                     "pred_decision shape must be "
-                     "(n_samples, n_classes) with "
-                     "multiclass target")
+    error_message = (r"The shape of pred_decision can not be 1d array"
+                     "with a multiclass target. "
+                     "pred_decision shape must be \(n_samples, n_classes\)")
+    with pytest.raises(ValueError, match=error_message):
+        hinge_loss(y_true=y_true, pred_decision=pred_decision)
 
-    assert_raise_message(ValueError, error_message, hinge_loss,
-                         y_true=y_true, pred_decision=pred_decision)
+    # test for inconsistency between pred_decision shape and labels number
     pred_decision = [[0, 1], [0, 1], [0, 1], [0, 1],
                      [2, 0], [0, 1], [1, 0]]
     labels = [0, 1, 2]
-    assert_raise_message(ValueError, error_message, hinge_loss,
-                         y_true=y_true, pred_decision=pred_decision,
-                         labels=labels)
+    error_message = (r"The shape of pred_decision is not "
+                     "consistent with the number of classes. "
+                     "pred_decision shape must be "
+                     "\(n_samples, n_classes\) with "
+                     "multiclass target")
+    with pytest.raises(ValueError, match=error_message):
+        hinge_loss(y_true=y_true, pred_decision=pred_decision, labels=labels)
 
 
 def test_hinge_loss_multiclass_with_missing_labels():

From 02ad7f75a7d97dfa3ffefc91cd48f282170732b1 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 09:16:11 +0100
Subject: [PATCH 07/16] line with less than 80 characters.

---
 doc/whats_new/v1.0.rst | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/doc/whats_new/v1.0.rst b/doc/whats_new/v1.0.rst
index 895d59a4d4fde..a8bb18dca2f2e 100644
--- a/doc/whats_new/v1.0.rst
+++ b/doc/whats_new/v1.0.rst
@@ -165,9 +165,9 @@ Changelog
   class methods and will be removed in 1.2.
   :pr:`18543` by `Guillaume Lemaitre`_.
 
-- |Enhancement| A fix to raise an error in :func:`metrics.hinge_loss` when ``pred_decision`` is 1d
-  whereas it is a multiclass classification or when ``pred_decision``
-  parameter is not consistent with ``labels`` parameter.
+- |Enhancement| A fix to raise an error in :func:`metrics.hinge_loss` when
+  ``pred_decision`` is 1d whereas it is a multiclass classification or when
+  ``pred_decision`` parameter is not consistent with ``labels`` parameter.
   :pr:`19643` by :user:`Pierre Attard <PierreAttard>`.
 
 - |Feature| :func:`metrics.mean_pinball_loss` exposes the pinball loss for

From c2937c2f2ebb97a1ec617068f0673775614e6ea5 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 09:24:59 +0100
Subject: [PATCH 08/16] Code standard modification E127 and E501

---
 sklearn/metrics/_classification.py           | 4 ++--
 sklearn/metrics/tests/test_classification.py | 3 ++-
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index 7fa3fd19ce616..0d6b4df89c171 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2310,8 +2310,8 @@ def _check_valid_multiclass_decision_shape(y_true_unique, pred_decision,
                          "with a multiclass target. "
                          "pred_decision shape must be (n_samples, n_classes)")
 
-    invalid_decision_shape = pred_decision.ndim > 1 and \
-                             y_true_unique.size != pred_decision.shape[1]
+    invalid_decision_shape = (pred_decision.ndim > 1 and
+                              y_true_unique.size != pred_decision.shape[1])
 
     if labels is None:
         if invalid_decision_shape:
diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 28c9524960bd4..61bb5758eb4da 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -2136,7 +2136,8 @@ def test_hinge_loss_multiclass_missing_labels_with_labels_none():
 
 
 def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
-    # test for inconsistency between multiclass problem and pred_decision argument
+    # test for inconsistency between multiclass problem and pred_decision
+    # argument
     y_true = [2, 1, 0, 1, 0, 1, 1]
     pred_decision = np.array([0, 1, 2, 1, 0, 2, 1])
     error_message = (r"The shape of pred_decision can not be 1d array"

From f2d5a7b240b44a373e9feea70c6829c7162f297c Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Thu, 11 Mar 2021 09:38:20 +0100
Subject: [PATCH 09/16] resolve W605 invalid escape sequence

---
 sklearn/metrics/tests/test_classification.py | 24 +++++++++-----------
 1 file changed, 11 insertions(+), 13 deletions(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 61bb5758eb4da..278769dcdffd5 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -1,4 +1,3 @@
-
 from functools import partial
 from itertools import product
 from itertools import chain
@@ -50,6 +49,7 @@
 
 from scipy.spatial.distance import hamming as sp_hamming
 
+
 ###############################################################################
 # Utilities for testing
 
@@ -102,7 +102,6 @@ def make_prediction(dataset=None, binary=False):
 # Tests
 
 def test_classification_report_dictionary_output():
-
     # Test performance report with dictionary output
     iris = datasets.load_iris()
     y_true, y_pred, _ = make_prediction(dataset=iris, binary=False)
@@ -135,7 +134,7 @@ def test_classification_report_dictionary_output():
         target_names=iris.target_names, output_dict=True)
 
     # assert the 2 dicts are equal.
-    assert(report.keys() == expected_report.keys())
+    assert (report.keys() == expected_report.keys())
     for key in expected_report:
         if key == 'accuracy':
             assert isinstance(report[key], float)
@@ -616,9 +615,9 @@ def test_cohen_kappa():
     y2 = np.array([0] * 50 + [1] * 40 + [2] * 10)
     assert_almost_equal(cohen_kappa_score(y1, y2), .9315, decimal=4)
     assert_almost_equal(cohen_kappa_score(y1, y2,
-                        weights="linear"), 0.9412, decimal=4)
+                                          weights="linear"), 0.9412, decimal=4)
     assert_almost_equal(cohen_kappa_score(y1, y2,
-                        weights="quadratic"), 0.9541, decimal=4)
+                                          weights="quadratic"), 0.9541, decimal=4)
 
 
 @ignore_warnings
@@ -782,7 +781,7 @@ def mcc_safe(y_true, y_pred):
         mcc_denominator = activity * pos_rate * (1 - activity) * (1 - pos_rate)
         return mcc_numerator / np.sqrt(mcc_denominator)
 
-    def random_ys(n_points):    # binary
+    def random_ys(n_points):  # binary
         x_true = rng.random_sample(n_points)
         x_pred = x_true + 0.2 * (rng.random_sample(n_points) - 0.5)
         y_true = (x_true > 0.5)
@@ -1237,7 +1236,7 @@ def test_multilabel_hamming_loss():
     assert hamming_loss(y1, np.zeros(y1.shape)) == 4 / 6
     assert hamming_loss(y2, np.zeros(y1.shape)) == 0.5
     assert hamming_loss(y1, y2, sample_weight=w) == 1. / 12
-    assert hamming_loss(y1, 1-y2, sample_weight=w) == 11. / 12
+    assert hamming_loss(y1, 1 - y2, sample_weight=w) == 11. / 12
     assert hamming_loss(y1, np.zeros_like(y1), sample_weight=w) == 2. / 3
     # sp_hamming only works with 1-D arrays
     assert hamming_loss(y1[0], y2[0]) == sp_hamming(y1[0], y2[0])
@@ -1436,6 +1435,7 @@ def test_jaccard_score_zero_division_set_value(zero_division, expected_score):
     assert score == pytest.approx(expected_score)
     assert len(record) == 0
 
+
 @ignore_warnings
 def test_precision_recall_f1_score_multilabel_1():
     # Test precision_recall_f1_score on a crafted multilabel example
@@ -1758,7 +1758,6 @@ def test_prf_warnings():
     # average of per-label scores
     f, w = precision_recall_fscore_support, UndefinedMetricWarning
     for average in [None, 'weighted', 'macro']:
-
         msg = ('Precision and F-score are ill-defined and '
                'being set to 0.0 in labels with no predicted samples.'
                ' Use `zero_division` parameter to control'
@@ -1834,7 +1833,6 @@ def test_prf_no_warnings_if_zero_division_set(zero_division):
     # average of per-label scores
     f = precision_recall_fscore_support
     for average in [None, 'weighted', 'macro']:
-
         assert_no_warnings(f, [0, 1, 2], [1, 1, 2], average=average,
                            zero_division=zero_division)
 
@@ -2140,9 +2138,9 @@ def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
     # argument
     y_true = [2, 1, 0, 1, 0, 1, 1]
     pred_decision = np.array([0, 1, 2, 1, 0, 2, 1])
-    error_message = (r"The shape of pred_decision can not be 1d array"
+    error_message = ("The shape of pred_decision can not be 1d array"
                      "with a multiclass target. "
-                     "pred_decision shape must be \(n_samples, n_classes\)")
+                     r"pred_decision shape must be \(n_samples, n_classes\)")
     with pytest.raises(ValueError, match=error_message):
         hinge_loss(y_true=y_true, pred_decision=pred_decision)
 
@@ -2150,10 +2148,10 @@ def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
     pred_decision = [[0, 1], [0, 1], [0, 1], [0, 1],
                      [2, 0], [0, 1], [1, 0]]
     labels = [0, 1, 2]
-    error_message = (r"The shape of pred_decision is not "
+    error_message = ("The shape of pred_decision is not "
                      "consistent with the number of classes. "
                      "pred_decision shape must be "
-                     "\(n_samples, n_classes\) with "
+                     r"\(n_samples, n_classes\) with "
                      "multiclass target")
     with pytest.raises(ValueError, match=error_message):
         hinge_loss(y_true=y_true, pred_decision=pred_decision, labels=labels)

From e30624e65cd2520cf4f15761d90c7b57101badbe Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Thu, 11 Mar 2021 09:41:34 +0100
Subject: [PATCH 10/16] resolve E501 line too long

---
 sklearn/metrics/tests/test_classification.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 278769dcdffd5..25d6cb80784dc 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -616,8 +616,8 @@ def test_cohen_kappa():
     assert_almost_equal(cohen_kappa_score(y1, y2), .9315, decimal=4)
     assert_almost_equal(cohen_kappa_score(y1, y2,
                                           weights="linear"), 0.9412, decimal=4)
-    assert_almost_equal(cohen_kappa_score(y1, y2,
-                                          weights="quadratic"), 0.9541, decimal=4)
+    assert_almost_equal(cohen_kappa_score(y1, y2, weights="quadratic"),
+                        0.9541, decimal=4)
 
 
 @ignore_warnings

From a01947c760c94d8cc3fb77cfd457d48e53ec6ebe Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Thu, 11 Mar 2021 17:22:00 +0100
Subject: [PATCH 11/16] remove function use only one time. 1d pred_decision for
 a multiclass terget.

---
 sklearn/metrics/_classification.py | 60 ++++++++----------------------
 1 file changed, 16 insertions(+), 44 deletions(-)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index 0d6b4df89c171..ee15f905c79c3 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2285,48 +2285,6 @@ def log_loss(y_true, y_pred, *, eps=1e-15, normalize=True, sample_weight=None,
     return _weighted_sum(loss, sample_weight, normalize)
 
 
-def _check_valid_multiclass_decision_shape(y_true_unique, pred_decision,
-                                           labels):
-    """Check if the inputs are consistent.
-
-    If it is a multiclass problem (it means thare are more than 2 classes
-    to classify), `pred_decision` parameter can't be 1d array. The column
-    number must be equal to classes number.
-
-    If not every classes are present in `y_true`, `labels` parameter
-    must be filled in.
-
-    Parameters
-    ----------
-    y_true_unique : array-like
-
-    pred_decision : array-like
-
-    labels : array-like
-    """
-
-    if pred_decision.ndim <= 1:
-        raise ValueError("The shape of pred_decision can not be 1d array"
-                         "with a multiclass target. "
-                         "pred_decision shape must be (n_samples, n_classes)")
-
-    invalid_decision_shape = (pred_decision.ndim > 1 and
-                              y_true_unique.size != pred_decision.shape[1])
-
-    if labels is None:
-        if invalid_decision_shape:
-            raise ValueError("Please include all labels in y_true "
-                             "or pass labels as third argument")
-    elif invalid_decision_shape:
-        raise ValueError("The shape of pred_decision is not "
-                         "consistent with the number of classes. "
-                         "pred_decision shape must be "
-                         "(n_samples, n_classes) with "
-                         "multiclass target")
-
-    return
-
-
 @_deprecate_positional_args
 def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     """Average hinge loss (non-regularized).
@@ -2415,8 +2373,22 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
 
     if y_true_unique.size > 2:
 
-        _check_valid_multiclass_decision_shape(y_true_unique, pred_decision,
-                                               labels)
+        if pred_decision.ndim <= 1:
+            raise ValueError("The shape of pred_decision can not be 1d array"
+                             "with a multiclass target. pred_decision shape "
+                             "must be (n_samples, n_classes)")
+
+        # pred_decision.ndim > 1 is true
+        if y_true_unique.size != pred_decision.shape[1]:
+            if labels is None:
+                raise ValueError("Please include all labels in y_true "
+                                 "or pass labels as third argument")
+            else:
+                raise ValueError("The shape of pred_decision is not "
+                                 "consistent with the number of classes. "
+                                 "pred_decision shape must be "
+                                 "(n_samples, n_classes) with "
+                                 "multiclass target")
         if labels is None:
             labels = y_true_unique
         le = LabelEncoder()

From 0d3b31e85d1a77932a45ecf92033acc558c7969a Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Thu, 11 Mar 2021 17:46:42 +0100
Subject: [PATCH 12/16] revert accidental cosmetic changes caused by an editor.

---
 sklearn/metrics/tests/test_classification.py | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 25d6cb80784dc..7cbc4dc562b39 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -1,3 +1,4 @@
+
 from functools import partial
 from itertools import product
 from itertools import chain
@@ -49,7 +50,6 @@
 
 from scipy.spatial.distance import hamming as sp_hamming
 
-
 ###############################################################################
 # Utilities for testing
 
@@ -102,6 +102,7 @@ def make_prediction(dataset=None, binary=False):
 # Tests
 
 def test_classification_report_dictionary_output():
+
     # Test performance report with dictionary output
     iris = datasets.load_iris()
     y_true, y_pred, _ = make_prediction(dataset=iris, binary=False)
@@ -134,7 +135,7 @@ def test_classification_report_dictionary_output():
         target_names=iris.target_names, output_dict=True)
 
     # assert the 2 dicts are equal.
-    assert (report.keys() == expected_report.keys())
+    assert(report.keys() == expected_report.keys())
     for key in expected_report:
         if key == 'accuracy':
             assert isinstance(report[key], float)
@@ -615,9 +616,9 @@ def test_cohen_kappa():
     y2 = np.array([0] * 50 + [1] * 40 + [2] * 10)
     assert_almost_equal(cohen_kappa_score(y1, y2), .9315, decimal=4)
     assert_almost_equal(cohen_kappa_score(y1, y2,
-                                          weights="linear"), 0.9412, decimal=4)
-    assert_almost_equal(cohen_kappa_score(y1, y2, weights="quadratic"),
-                        0.9541, decimal=4)
+                        weights="linear"), 0.9412, decimal=4)
+    assert_almost_equal(cohen_kappa_score(y1, y2,
+                        weights="quadratic"), 0.9541, decimal=4)
 
 
 @ignore_warnings
@@ -781,7 +782,7 @@ def mcc_safe(y_true, y_pred):
         mcc_denominator = activity * pos_rate * (1 - activity) * (1 - pos_rate)
         return mcc_numerator / np.sqrt(mcc_denominator)
 
-    def random_ys(n_points):  # binary
+    def random_ys(n_points):    # binary
         x_true = rng.random_sample(n_points)
         x_pred = x_true + 0.2 * (rng.random_sample(n_points) - 0.5)
         y_true = (x_true > 0.5)
@@ -1236,7 +1237,7 @@ def test_multilabel_hamming_loss():
     assert hamming_loss(y1, np.zeros(y1.shape)) == 4 / 6
     assert hamming_loss(y2, np.zeros(y1.shape)) == 0.5
     assert hamming_loss(y1, y2, sample_weight=w) == 1. / 12
-    assert hamming_loss(y1, 1 - y2, sample_weight=w) == 11. / 12
+    assert hamming_loss(y1, 1-y2, sample_weight=w) == 11. / 12
     assert hamming_loss(y1, np.zeros_like(y1), sample_weight=w) == 2. / 3
     # sp_hamming only works with 1-D arrays
     assert hamming_loss(y1[0], y2[0]) == sp_hamming(y1[0], y2[0])
@@ -1435,7 +1436,6 @@ def test_jaccard_score_zero_division_set_value(zero_division, expected_score):
     assert score == pytest.approx(expected_score)
     assert len(record) == 0
 
-
 @ignore_warnings
 def test_precision_recall_f1_score_multilabel_1():
     # Test precision_recall_f1_score on a crafted multilabel example
@@ -1758,6 +1758,7 @@ def test_prf_warnings():
     # average of per-label scores
     f, w = precision_recall_fscore_support, UndefinedMetricWarning
     for average in [None, 'weighted', 'macro']:
+
         msg = ('Precision and F-score are ill-defined and '
                'being set to 0.0 in labels with no predicted samples.'
                ' Use `zero_division` parameter to control'
@@ -1833,6 +1834,7 @@ def test_prf_no_warnings_if_zero_division_set(zero_division):
     # average of per-label scores
     f = precision_recall_fscore_support
     for average in [None, 'weighted', 'macro']:
+
         assert_no_warnings(f, [0, 1, 2], [1, 1, 2], average=average,
                            zero_division=zero_division)
 

From 4b8839e0173fbabb35e9e1a60f5a12cdf9afc379 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 19:44:49 +0100
Subject: [PATCH 13/16] Update doc/whats_new/v1.0.rst

Co-authored-by: Olivier Grisel <olivier.grisel@ensta.org>
---
 doc/whats_new/v1.0.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/whats_new/v1.0.rst b/doc/whats_new/v1.0.rst
index a8bb18dca2f2e..06764d2be6003 100644
--- a/doc/whats_new/v1.0.rst
+++ b/doc/whats_new/v1.0.rst
@@ -167,7 +167,7 @@ Changelog
 
 - |Enhancement| A fix to raise an error in :func:`metrics.hinge_loss` when
   ``pred_decision`` is 1d whereas it is a multiclass classification or when
-  ``pred_decision`` parameter is not consistent with ``labels`` parameter.
+  ``pred_decision`` parameter is not consistent with the ``labels`` parameter.
   :pr:`19643` by :user:`Pierre Attard <PierreAttard>`.
 
 - |Feature| :func:`metrics.mean_pinball_loss` exposes the pinball loss for

From 9bd35c239662ba63cd53a924b554944f4018fb58 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Thu, 11 Mar 2021 21:22:08 +0100
Subject: [PATCH 14/16] Add precision to error messages.

---
 sklearn/metrics/_classification.py           | 14 +++++++----
 sklearn/metrics/tests/test_classification.py | 26 +++++++++++---------
 2 files changed, 24 insertions(+), 16 deletions(-)

diff --git a/sklearn/metrics/_classification.py b/sklearn/metrics/_classification.py
index ee15f905c79c3..97ee5a2e01340 100644
--- a/sklearn/metrics/_classification.py
+++ b/sklearn/metrics/_classification.py
@@ -2374,9 +2374,11 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
     if y_true_unique.size > 2:
 
         if pred_decision.ndim <= 1:
-            raise ValueError("The shape of pred_decision can not be 1d array"
+            raise ValueError("The shape of pred_decision cannot be 1d array"
                              "with a multiclass target. pred_decision shape "
-                             "must be (n_samples, n_classes)")
+                             "must be (n_samples, n_classes), that is "
+                             f"({y_true.shape[0]}, {y_true_unique.size})."
+                             f" Got: {pred_decision.shape}")
 
         # pred_decision.ndim > 1 is true
         if y_true_unique.size != pred_decision.shape[1]:
@@ -2386,9 +2388,11 @@ def hinge_loss(y_true, pred_decision, *, labels=None, sample_weight=None):
             else:
                 raise ValueError("The shape of pred_decision is not "
                                  "consistent with the number of classes. "
-                                 "pred_decision shape must be "
-                                 "(n_samples, n_classes) with "
-                                 "multiclass target")
+                                 "With a multiclass target, pred_decision "
+                                 "shape must be "
+                                 "(n_samples, n_classes), that is "
+                                 f"({y_true.shape[0]}, {y_true_unique.size}). "
+                                 f"Got: {pred_decision.shape}")
         if labels is None:
             labels = y_true_unique
         le = LabelEncoder()
diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 7cbc4dc562b39..2925bcdc6bad4 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -4,6 +4,7 @@
 from itertools import chain
 from itertools import permutations
 import warnings
+import re
 
 import numpy as np
 from scipy import linalg
@@ -2138,24 +2139,27 @@ def test_hinge_loss_multiclass_missing_labels_with_labels_none():
 def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
     # test for inconsistency between multiclass problem and pred_decision
     # argument
-    y_true = [2, 1, 0, 1, 0, 1, 1]
+    y_true = np.array([2, 1, 0, 1, 0, 1, 1])
     pred_decision = np.array([0, 1, 2, 1, 0, 2, 1])
-    error_message = ("The shape of pred_decision can not be 1d array"
-                     "with a multiclass target. "
-                     r"pred_decision shape must be \(n_samples, n_classes\)")
-    with pytest.raises(ValueError, match=error_message):
+    error_message = ("The shape of pred_decision cannot be 1d array"
+                     "with a multiclass target. pred_decision shape "
+                     "must be (n_samples, n_classes), that is "
+                     f"({y_true.shape[0]}, {np.unique(y_true).size}). "
+                     f"Got: ({y_true.shape[0]},)")
+    with pytest.raises(ValueError, match=re.escape(error_message)):
         hinge_loss(y_true=y_true, pred_decision=pred_decision)
 
     # test for inconsistency between pred_decision shape and labels number
-    pred_decision = [[0, 1], [0, 1], [0, 1], [0, 1],
-                     [2, 0], [0, 1], [1, 0]]
+    pred_decision = np.array([[0, 1], [0, 1], [0, 1], [0, 1],
+                              [2, 0], [0, 1], [1, 0]])
     labels = [0, 1, 2]
     error_message = ("The shape of pred_decision is not "
                      "consistent with the number of classes. "
-                     "pred_decision shape must be "
-                     r"\(n_samples, n_classes\) with "
-                     "multiclass target")
-    with pytest.raises(ValueError, match=error_message):
+                     "With a multiclass target, pred_decision "
+                     "shape must be (n_samples, n_classes), that is "
+                     f"({y_true.shape[0]}, {np.unique(y_true).size}). "
+                     f"Got: ({y_true.shape[0]}, {pred_decision.shape[1]})")
+    with pytest.raises(ValueError, match=re.escape(error_message)):
         hinge_loss(y_true=y_true, pred_decision=pred_decision, labels=labels)
 
 

From fd2f658db8e6b2e257fafa3e26c562e56fb9c0b7 Mon Sep 17 00:00:00 2001
From: PierreAttard <attard.pierre@gmail.com>
Date: Fri, 12 Mar 2021 18:37:43 +0100
Subject: [PATCH 15/16] Update sklearn/metrics/tests/test_classification.py

Co-authored-by: Olivier Grisel <olivier.grisel@ensta.org>
---
 sklearn/metrics/tests/test_classification.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 2925bcdc6bad4..335ae5f8dd780 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -2145,7 +2145,7 @@ def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
                      "with a multiclass target. pred_decision shape "
                      "must be (n_samples, n_classes), that is "
                      f"({y_true.shape[0]}, {np.unique(y_true).size}). "
-                     f"Got: ({y_true.shape[0]},)")
+                     f"Got: {pred_decision.shape}")
     with pytest.raises(ValueError, match=re.escape(error_message)):
         hinge_loss(y_true=y_true, pred_decision=pred_decision)
 

From 6cb0dad2896675077ba60abce910707a0a80c562 Mon Sep 17 00:00:00 2001
From: "p.attard" <Eay1zdefr38*$>
Date: Fri, 12 Mar 2021 18:41:50 +0100
Subject: [PATCH 16/16] hard code expected and observed values in unitest

---
 sklearn/metrics/tests/test_classification.py | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/sklearn/metrics/tests/test_classification.py b/sklearn/metrics/tests/test_classification.py
index 2925bcdc6bad4..7b634e88f2275 100644
--- a/sklearn/metrics/tests/test_classification.py
+++ b/sklearn/metrics/tests/test_classification.py
@@ -2144,8 +2144,7 @@ def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
     error_message = ("The shape of pred_decision cannot be 1d array"
                      "with a multiclass target. pred_decision shape "
                      "must be (n_samples, n_classes), that is "
-                     f"({y_true.shape[0]}, {np.unique(y_true).size}). "
-                     f"Got: ({y_true.shape[0]},)")
+                     "(7, 3). Got: (7,)")
     with pytest.raises(ValueError, match=re.escape(error_message)):
         hinge_loss(y_true=y_true, pred_decision=pred_decision)
 
@@ -2157,8 +2156,7 @@ def test_hinge_loss_multiclass_no_consistent_pred_decision_shape():
                      "consistent with the number of classes. "
                      "With a multiclass target, pred_decision "
                      "shape must be (n_samples, n_classes), that is "
-                     f"({y_true.shape[0]}, {np.unique(y_true).size}). "
-                     f"Got: ({y_true.shape[0]}, {pred_decision.shape[1]})")
+                     "(7, 3). Got: (7, 2)")
     with pytest.raises(ValueError, match=re.escape(error_message)):
         hinge_loss(y_true=y_true, pred_decision=pred_decision, labels=labels)