[MRG] Check for NaN or Inf in metric input (scikit-learn#6976)

Author: betatim

sklearn/metrics/classification.py

@@ -30,6 +30,7 @@
 
 from ..preprocessing import LabelBinarizer, label_binarize
 from ..preprocessing import LabelEncoder
+from ..utils import assert_all_finite
 from ..utils import check_array
 from ..utils import check_consistent_length
 from ..utils import column_or_1d
@@ -1843,6 +1844,9 @@ def brier_score_loss(y_true, y_prob, sample_weight=None, pos_label=None):
     """
     y_true = column_or_1d(y_true)
     y_prob = column_or_1d(y_prob)
+    assert_all_finite(y_true)
+    assert_all_finite(y_prob)
+
     if pos_label is None:
         pos_label = y_true.max()
     y_true = np.array(y_true == pos_label, int)

sklearn/metrics/ranking.py

@@ -23,6 +23,7 @@
 import numpy as np
 from scipy.sparse import csr_matrix
 
+from ..utils import assert_all_finite
 from ..utils import check_consistent_length
 from ..utils import column_or_1d, check_array
 from ..utils.multiclass import type_of_target
@@ -296,6 +297,9 @@ def _binary_clf_curve(y_true, y_score, pos_label=None, sample_weight=None):
     check_consistent_length(y_true, y_score)
     y_true = column_or_1d(y_true)
     y_score = column_or_1d(y_score)
+    assert_all_finite(y_true)
+    assert_all_finite(y_score)
+
     if sample_weight is not None:
         sample_weight = column_or_1d(sample_weight)
 

sklearn/metrics/tests/test_common.py

@@ -19,6 +19,7 @@
 from sklearn.utils.testing import assert_greater
 from sklearn.utils.testing import assert_not_equal
 from sklearn.utils.testing import assert_raises
+from sklearn.utils.testing import assert_raise_message
 from sklearn.utils.testing import assert_true
 from sklearn.utils.testing import ignore_warnings
 
@@ -608,6 +609,29 @@ def test_invariance_string_vs_numbers_labels():
             assert_raises(ValueError, metric, y1_str.astype('O'), y2)
 
 
+def test_inf_nan_input():
+    invalids =[([0, 1], [np.inf, np.inf]),
+               ([0, 1], [np.nan, np.nan]),
+               ([0, 1], [np.nan, np.inf])]
+
+    METRICS = dict()
+    METRICS.update(THRESHOLDED_METRICS)
+    METRICS.update(REGRESSION_METRICS)
+
+    for metric in METRICS.values():
+        for y_true, y_score in invalids:
+            assert_raise_message(ValueError,
+                                 "contains NaN, infinity",
+                                 metric, y_true, y_score)
+
+    # Classification metrics all raise a mixed input exception
+    for metric in CLASSIFICATION_METRICS.values():
+        for y_true, y_score in invalids:
+            assert_raise_message(ValueError,
+                                 "Can't handle mix of binary and continuous",
+                                 metric, y_true, y_score)
+
+
 @ignore_warnings
 def check_single_sample(name):
     # Non-regression test: scores should work with a single sample.