scikit-learn · aishgrt1 · Apr 15, 2018 · Apr 15, 2018 · Apr 15, 2018 · Apr 15, 2018
diff --git a/doc/modules/calibration.rst b/doc/modules/calibration.rst
@@ -111,7 +111,8 @@ redundant. The figure shows the estimated probabilities obtained with
 logistic regression, a linear support-vector classifier (SVC), and linear SVC with
 both isotonic calibration and sigmoid calibration. The calibration performance
 is evaluated with Brier score :func:`brier_score_loss`, reported in the legend
-(the smaller the better).
+(the smaller the better). The Brier Score is composed of calibration loss and
+refinment loss.
 
 .. figure:: ../auto_examples/calibration/images/sphx_glr_plot_calibration_curve_002.png
    :target: ../auto_examples/calibration/plot_calibration_curve.html

diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
@@ -141,6 +141,15 @@ Classifiers and regressors
 - :class:`dummy.DummyClassifier` and :class:`dummy.DummyRegresssor` now
   only require X to be an object with finite length or shape.
   :issue:`9832` by :user:`Vrishank Bhardwaj <vrishank97>`.
+
+- :class:`neighbors.RadiusNeighborsRegressor` and
+  :class:`neighbors.RadiusNeighborsClassifier` are now
+  parallelized according to ``n_jobs`` regardless of ``algorithm``.
+  :issue:`8003` by :user:`Joël Billaud <recamshak>`.
+
+- Clipped the values of proba between [0,1] in
+  :class:`calibration` for probability values to lie between [0,1].
+  :issue: `10903` by :user:`Aishwarya Srinivasan <aishgrt1>`.
 
 Cluster
 

diff --git a/sklearn/calibration.py b/sklearn/calibration.py
@@ -388,8 +388,8 @@ def predict_proba(self, X):
         # XXX : for some reason all probas can be 0
         proba[np.isnan(proba)] = 1. / n_classes
 
-        # Deal with cases where the predicted probability minimally exceeds 1.0
-        proba[(1.0 < proba) & (proba <= 1.0 + 1e-5)] = 1.0
+        proba[proba > 1.0] = 1.0
+        proba[proba < 0.0] = 0.0
 
         return proba
 

diff --git a/sklearn/tests/data_X_test.pickle b/sklearn/tests/data_X_test.pickle
diff --git a/sklearn/tests/data_X_train.pickle b/sklearn/tests/data_X_train.pickle
diff --git a/sklearn/tests/data_y_train.pickle b/sklearn/tests/data_y_train.pickle
diff --git a/sklearn/tests/test_calibration.py b/sklearn/tests/test_calibration.py
@@ -4,6 +4,7 @@
 from __future__ import division
 import numpy as np
 from scipy import sparse
+import pickle
 from sklearn.model_selection import LeaveOneOut
 
 from sklearn.utils.testing import (assert_array_almost_equal, assert_equal,
@@ -13,7 +14,7 @@
                                    assert_raises,
                                    ignore_warnings)
 from sklearn.datasets import make_classification, make_blobs
-from sklearn.naive_bayes import MultinomialNB
+from sklearn.naive_bayes import MultinomialNB, GaussianNB
 from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
 from sklearn.svm import LinearSVC
 from sklearn.pipeline import Pipeline
@@ -239,6 +240,26 @@ def test_sigmoid_calibration():
                   np.vstack((exF, exF)), exY)
 
 
+def test_isotonic_calibration_prob_not_inf():
+    # Test from @LotusZephyr's issue:
+    # https://github.com/scikit-learn/scikit-learn/issues/10903
+    with open('data_X_train.pickle', 'rb') as f:
+        X_train = pickle.load(f)
+
+    with open('data_X_test.pickle', 'rb') as f:
+        X_test = pickle.load(f)
+
+    with open('data_y_train.pickle', 'rb') as f:
+        y_train = pickle.load(f)
+
+    clf = GaussianNB()
+    clf_c = CalibratedClassifierCV(clf, method='isotonic')
+    clf_fit = clf_c.fit(X_train, y_train)
+    y_pred = clf_fit.predict_proba(X_test)[:, 1]
+    assert(np.all(y_pred >= 0))
+    assert(np.all(y_pred <= 1))
+
+
 def test_calibration_curve():
     """Check calibration_curve function"""
     y_true = np.array([0, 0, 0, 1, 1, 1])