FIX Use take instead of choose in compute_sample_weight (#12165)

scikit-learn · Oct 15, 2018 · e2a7b31 · e2a7b31
1 parent cec0fba
commit e2a7b31
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 6 deletions.
diff --git a/doc/whats_new/v0.20.rst b/doc/whats_new/v0.20.rst
@@ -17,6 +17,10 @@ enhancements to features released in 0.20.0.
   ``n_jobs > 1``.
   :issue:`12159` by :user:`Olivier Grisel <ogrisel>`.
 
+- |Fix| Fixed a bug mostly affecting :class:`ensemble.RandomForestClassifier`
+  where ``class_weight='balanced_subsample'`` failed with more than 32 classes.
+  :issue:`12165` by `Joel Nothman`_.
+
 - |Fix| :func:`linear_model.SGDClassifier` and variants
   with ``early_stopping=True`` would not use a consistent validation
   split in the multiclass case and this would cause a crash when using

diff --git a/sklearn/utils/class_weight.py b/sklearn/utils/class_weight.py
@@ -150,12 +150,12 @@ def compute_sample_weight(class_weight, y, indices=None):
             y_subsample = y[indices, k]
             classes_subsample = np.unique(y_subsample)
 
-            weight_k = np.choose(np.searchsorted(classes_subsample,
-                                                 classes_full),
-                                 compute_class_weight(class_weight_k,
-                                                      classes_subsample,
-                                                      y_subsample),
-                                 mode='clip')
+            weight_k = np.take(compute_class_weight(class_weight_k,
+                                                    classes_subsample,
+                                                    y_subsample),
+                               np.searchsorted(classes_subsample,
+                                               classes_full),
+                               mode='clip')
 
             classes_missing = set(classes_full) - set(classes_subsample)
         else:

diff --git a/sklearn/utils/tests/test_class_weight.py b/sklearn/utils/tests/test_class_weight.py
@@ -251,3 +251,11 @@ def test_compute_sample_weight_errors():
 
     # Incorrect length list for multi-output
     assert_raises(ValueError, compute_sample_weight, [{1: 2, 2: 1}], y_)
+
+
+def test_compute_sample_weight_more_than_32():
+    # Non-regression smoke test for #12146
+    y = np.arange(50)  # more than 32 distinct classes
+    indices = np.arange(50)  # use subsampling
+    weight = compute_sample_weight('balanced', y, indices=indices)
+    assert_array_almost_equal(weight, np.ones(y.shape[0]))