cross_validation

scikit-learn · Apr 22, 2014 · 0ce2180 · 0ce2180
1 parent c5867a7
commit 0ce2180
Showing 1 changed file with 8 additions and 11 deletions.
diff --git a/sklearn/cross_validation.py b/sklearn/cross_validation.py
@@ -1115,11 +1115,11 @@ def cross_val_score(estimator, X, y=None, sample_weight=None,
     # independent, and that it is pickle-able.
     parallel = Parallel(n_jobs=n_jobs, verbose=verbose,
                         pre_dispatch=pre_dispatch)
-    scores = parallel(delayed(_fit_and_score)(
-        clone(estimator), X, y, sample_weight, scorer,
-        train, test, verbose, None,
-        fit_params)
-        for train, test in cv)
+    scores = parallel(delayed(_fit_and_score)(clone(estimator), X, y,
+                                              sample_weight, scorer,
+                                              train, test, verbose, None,
+                                              fit_params)
+                      for train, test in cv)
     return np.array(scores)[:, 0]
 
 
@@ -1211,8 +1211,8 @@ def _fit_and_score(estimator, X, y, sample_weight,
     X_test, y_test, sample_weight_test = _safe_split(
         estimator, X, y, sample_weight, test, train)
 
-    test_score_params = dict()
-    train_score_params = dict()
+    test_score_params = {}
+    train_score_params = {}
     if sample_weight is not None:
         fit_params = fit_params.copy()
         fit_params['sample_weight'] = sample_weight_train
@@ -1274,10 +1274,7 @@ def _safe_split(estimator, X, y, sample_weight, indices, train_indices=None):
         y_subset = None
 
     if sample_weight is not None:
-        if not hasattr(sample_weight, "shape"):
-            sample_weight_subset = [sample_weight[idx] for idx in indices]
-        else:
-            sample_weight_subset = sample_weight[indices]
+        sample_weight_subset = np.asarray(sample_weight)[indices]
     else:
         sample_weight_subset = None