Fixes to sparse decision_function, rebase fixes.

Author: amueller

doc/whats_new.rst

@@ -207,6 +207,9 @@ Enhancements
    - Numerical stability improvements for :class:`preprocessing.StandardScaler`
      and :func:`preprocessing.scale`. By `Nicolas Goix`_
 
+   - :class:`svm.SVC` fitted on sparse input now implements ``decision_function``.
+     By `Rob Zinkov`_ and `Andreas Müller`_.
+
 Documentation improvements
 ..........................
 

sklearn/svm/base.py

@@ -368,17 +368,15 @@ def decision_function(self, X):
         X = self._validate_for_predict(X)
         X = self._compute_kernel(X)
 
-        X = self._validate_for_predict(X)
-
         if self._sparse:
             dec_func = self._sparse_decision_function(X)
         else:
             dec_func = self._dense_decision_function(X)
 
         # In binary case, we need to flip the sign of coef, intercept and
         # decision function.
-        if self.impl != 'one_class' and len(self.classes_) == 2:
-            return -dec_func
+        if self._impl in ['c_svc', 'nu_svc'] and len(self.classes_) == 2:
+            return -dec_func.ravel()
 
         return dec_func
 
@@ -412,11 +410,11 @@ def _sparse_decision_function(self, X):
             self.support_vectors_.indices,
             self.support_vectors_.indptr,
             self._dual_coef_.data, self._intercept_,
-            LIBSVM_IMPL.index(self.impl), kernel_type,
-            self.degree, self.gamma, self.coef0, self.tol,
+            LIBSVM_IMPL.index(self._impl), kernel_type,
+            self.degree, self._gamma, self.coef0, self.tol,
             self.C, self.class_weight_,
             self.nu, self.epsilon, self.shrinking,
-            self.probability, self.n_support_, self._label,
+            self.probability, self.n_support_,
             self.probA_, self.probB_)
 
     def _validate_for_predict(self, X):