Merge branch 'master' into pls_float_preserve

Mariand012 · web-flow · commit 508d95e2ef2c · 2018-04-22T13:20:53.000+01:00
diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
@@ -91,9 +91,12 @@ Preprocessing
 
 - Added :class:`MICEImputer`, which is a strategy for imputing missing
   values by modeling each feature with missing values as a function of
-  other features in a round-robin fashion. :issue:`8478` by 
+  other features in a round-robin fashion. :issue:`8478` by
   :user:`Sergey Feldman <sergeyf>`.
 
+- Updated :class:`preprocessing.MinMaxScaler` to pass through NaN values. :issue:`10404`
+  by :user:`Lucija Gregov <LucijaGregov>`.
+
 Model evaluation
 
 - Added the :func:`metrics.balanced_accuracy_score` metric and a corresponding
@@ -411,6 +414,12 @@ Preprocessing
   ``inverse_transform`` on unseen labels. :issue:`9816` by :user:`Charlie Newey
   <newey01c>`.
 
+Model evaluation and meta-estimators
+
+- Add improved error message in :func:`model_selection.cross_val_score` when
+  multiple metrics are passed in ``scoring`` keyword.
+  :issue:`11006` by :user:`Ming Li <minggli>`.
+
 Datasets
 
 - Fixed a bug in :func:`dataset.load_boston` which had a wrong data point.
diff --git a/sklearn/metrics/scorer.py b/sklearn/metrics/scorer.py
@@ -19,6 +19,7 @@
 # License: Simplified BSD
 
 from abc import ABCMeta, abstractmethod
+from collections import Iterable
 import warnings
 
 import numpy as np
@@ -300,6 +301,10 @@ def check_scoring(estimator, scoring=None, allow_none=False):
                 "If no scoring is specified, the estimator passed should "
                 "have a 'score' method. The estimator %r does not."
                 % estimator)
+    elif isinstance(scoring, Iterable):
+        raise ValueError("For evaluating multiple scores, use "
+                         "sklearn.model_selection.cross_validate instead. "
+                         "{0} was passed.".format(scoring))
     else:
         raise ValueError("scoring value should either be a callable, string or"
                          " None. %r was passed" % scoring)
diff --git a/sklearn/model_selection/tests/test_search.py b/sklearn/model_selection/tests/test_search.py
@@ -1217,10 +1217,10 @@ def test_fit_grid_point():
             assert_equal(n_test_samples, test.size)
 
     # Should raise an error upon multimetric scorer
-    assert_raise_message(ValueError, "scoring value should either be a "
-                         "callable, string or None.", fit_grid_point, X, y,
-                         svc, params, train, test, {'score': scorer},
-                         verbose=True)
+    assert_raise_message(ValueError, "For evaluating multiple scores, use "
+                         "sklearn.model_selection.cross_validate instead.",
+                         fit_grid_point, X, y, svc, params, train, test,
+                         {'score': scorer}, verbose=True)
 
 
 def test_pickle():
diff --git a/sklearn/preprocessing/data.py b/sklearn/preprocessing/data.py
@@ -276,6 +276,9 @@ class MinMaxScaler(BaseEstimator, TransformerMixin):
 
     Notes
     -----
+    NaNs are treated as missing values: disregarded in fit, and maintained in
+    transform.
+
     For a comparison of the different scalers, transformers, and normalizers,
     see :ref:`examples/preprocessing/plot_all_scaling.py
     <sphx_glr_auto_examples_preprocessing_plot_all_scaling.py>`.
@@ -340,10 +343,11 @@ def partial_fit(self, X, y=None):
                             "You may consider to use MaxAbsScaler instead.")
 
         X = check_array(X, copy=self.copy, warn_on_dtype=True,
-                        estimator=self, dtype=FLOAT_DTYPES)
+                        estimator=self, dtype=FLOAT_DTYPES,
+                        force_all_finite="allow-nan")
 
-        data_min = np.min(X, axis=0)
-        data_max = np.max(X, axis=0)
+        data_min = np.nanmin(X, axis=0)
+        data_max = np.nanmax(X, axis=0)
 
         # First pass
         if not hasattr(self, 'n_samples_seen_'):
@@ -373,7 +377,8 @@ def transform(self, X):
         """
         check_is_fitted(self, 'scale_')
 
-        X = check_array(X, copy=self.copy, dtype=FLOAT_DTYPES)
+        X = check_array(X, copy=self.copy, dtype=FLOAT_DTYPES,
+                        force_all_finite="allow-nan")
 
         X *= self.scale_
         X += self.min_
@@ -389,7 +394,8 @@ def inverse_transform(self, X):
         """
         check_is_fitted(self, 'scale_')
 
-        X = check_array(X, copy=self.copy, dtype=FLOAT_DTYPES)
+        X = check_array(X, copy=self.copy, dtype=FLOAT_DTYPES,
+                        force_all_finite="allow-nan")
 
         X -= self.min_
         X /= self.scale_
diff --git a/sklearn/preprocessing/tests/test_common.py b/sklearn/preprocessing/tests/test_common.py
@@ -4,6 +4,7 @@
 from sklearn.datasets import load_iris
 from sklearn.model_selection import train_test_split
 from sklearn.preprocessing import QuantileTransformer
+from sklearn.preprocessing import MinMaxScaler
 from sklearn.utils.testing import assert_array_equal
 from sklearn.utils.testing import assert_allclose
 
@@ -12,7 +13,8 @@
 
 @pytest.mark.parametrize(
     "est",
-    [QuantileTransformer(n_quantiles=10, random_state=42)]
+    [MinMaxScaler(),
+     QuantileTransformer(n_quantiles=10, random_state=42)]
 )
 def test_missing_value_handling(est):
     # check that the preprocessing method let pass nan
diff --git a/sklearn/utils/estimator_checks.py b/sklearn/utils/estimator_checks.py
@@ -73,7 +73,8 @@
                 'RANSACRegressor', 'RadiusNeighborsRegressor',
                 'RandomForestRegressor', 'Ridge', 'RidgeCV']
 
-ALLOW_NAN = ['QuantileTransformer', 'Imputer', 'SimpleImputer', 'MICEImputer']
+ALLOW_NAN = ['Imputer', 'SimpleImputer', 'MICEImputer',
+             'MinMaxScaler', 'QuantileTransformer']
 SKIP_DTYPE_CONVERSION_CHECK = [
              'AdditiveChi2Sampler', 'BernoulliRBM', 'Binarizer', 'Birch',
              'CCA', 'DictionaryLearning', 'FactorAnalysis', 'FastICA',
