COSMIT rm deprecated stuff -- lots of it

larsmans · larsmans · commit 77e69edcec4d · 2012-09-20T21:40:33.000+02:00
diff --git a/doc/whats_new.rst b/doc/whats_new.rst
@@ -16,6 +16,21 @@ Changelog
 
    - Speed up of :func:`metrics.precision_recall_curve` by Conrad Lee.
 
+API changes summary
+-------------------
+
+   - The module ``sklearn.linear_model.sparse`` is gone. Sparse matrix support
+     was already integrated into the "regular" linear models.
+
+   - ``sklearn.metrics.mean_square_error``, which incorrectly returned the
+     cumulated error, was removed. Use ``mean_squared_error`` instead.
+
+   - Passing ``class_weight`` parameters to ``fit`` methods is no longer
+     supported. Pass them to estimator constuctors instead.
+
+   - GMMs no longer have ``decode`` and ``rvs`` methods. Use the ``score``,
+     ``predict`` or ``sample`` methods instead.
+
 
 .. _changes_0_12:
 
diff --git a/sklearn/grid_search.py b/sklearn/grid_search.py
@@ -329,10 +329,10 @@ def __init__(self, estimator, param_grid, loss_func=None, score_func=None,
         self.pre_dispatch = pre_dispatch
 
     def _set_methods(self):
-        if hasattr(self._best_estimator_, 'predict'):
-            self.predict = self._best_estimator_.predict
-        if hasattr(self._best_estimator_, 'predict_proba'):
-            self.predict_proba = self._best_estimator_.predict_proba
+        if hasattr(self.best_estimator_, 'predict'):
+            self.predict = self.best_estimator_.predict
+        if hasattr(self.best_estimator_, 'predict_proba'):
+            self.predict_proba = self.best_estimator_.predict_proba
 
     def fit(self, X, y=None, **params):
         """Run fit with all sets of parameters
@@ -379,7 +379,7 @@ def _fit(self, X, y):
             params = next(iter(grid))
             base_clf.set_params(**params)
             base_clf.fit(X, y)
-            self._best_estimator_ = base_clf
+            self.best_estimator_ = base_clf
             self._set_methods()
             return self
 
@@ -434,7 +434,7 @@ def _fit(self, X, y):
             # clone first to work around broken estimators
             best_estimator = clone(base_clf).set_params(**best_params)
             best_estimator.fit(X, y, **self.fit_params)
-            self._best_estimator_ = best_estimator
+            self.best_estimator_ = best_estimator
             self._set_methods()
 
         # Store the computed scores
@@ -455,14 +455,3 @@ def score(self, X, y=None):
                              % self.best_estimator_)
         y_predicted = self.predict(X)
         return self.score_func(y, y_predicted)
-
-    # TODO around 0.13: remove this property, make it an attribute
-    @property
-    def best_estimator_(self):
-        if hasattr(self, '_best_estimator_'):
-            return self._best_estimator_
-        else:
-            raise RuntimeError("Grid search has to be run with 'refit=True'"
-                " to make predictions or obtain an instance  of the best "
-                " estimator. To obtain the best parameter settings, "
-                " use ``best_params_``.")
diff --git a/sklearn/linear_model/__init__.py b/sklearn/linear_model/__init__.py
@@ -26,7 +26,6 @@
 from .perceptron import Perceptron
 from .randomized_l1 import RandomizedLasso, RandomizedLogisticRegression, \
                            lasso_stability_path
-from . import sparse
 from .isotonic_regression_ import IsotonicRegression
 
 __all__ = ['ARDRegression',
@@ -65,5 +64,4 @@
            'lasso_stability_path',
            'orthogonal_mp',
            'orthogonal_mp_gram',
-           'ridge_regression',
-           'sparse']
+           'ridge_regression']
diff --git a/sklearn/linear_model/sparse/__init__.py b/sklearn/linear_model/sparse/__init__.py
diff --git a/sklearn/linear_model/sparse/coordinate_descent.py b/sklearn/linear_model/sparse/coordinate_descent.py
diff --git a/sklearn/linear_model/sparse/logistic.py b/sklearn/linear_model/sparse/logistic.py
diff --git a/sklearn/linear_model/sparse/stochastic_gradient.py b/sklearn/linear_model/sparse/stochastic_gradient.py
diff --git a/sklearn/linear_model/stochastic_gradient.py b/sklearn/linear_model/stochastic_gradient.py
@@ -450,8 +450,7 @@ def _partial_fit(self, X, y, n_iter, classes=None, sample_weight=None,
 
         return self
 
-    def partial_fit(self, X, y, classes=None,
-                    class_weight=None, sample_weight=None):
+    def partial_fit(self, X, y, classes=None, sample_weight=None):
         """Fit linear model with Stochastic Gradient Descent.
 
         Parameters
@@ -478,12 +477,6 @@ def partial_fit(self, X, y, classes=None,
         -------
         self : returns an instance of self.
         """
-        if class_weight is not None:
-            warnings.warn("Using 'class_weight' as a parameter to the 'fit'"
-                          "method is deprecated and will be removed in 0.13. "
-                          "Set it on initialization instead.",
-                          DeprecationWarning, stacklevel=2)
-            self.class_weight = class_weight
         return self._partial_fit(X, y, n_iter=1, classes=classes,
                                  sample_weight=sample_weight)
 
diff --git a/sklearn/metrics/__init__.py b/sklearn/metrics/__init__.py
@@ -3,12 +3,12 @@
 and pairwise metrics and distance computations.
 """
 
-from .metrics import confusion_matrix, roc_curve, auc, precision_score, \
-                recall_score, fbeta_score, f1_score, zero_one_score, \
-                precision_recall_fscore_support, classification_report, \
-                precision_recall_curve, explained_variance_score, r2_score, \
-                zero_one, mean_square_error, hinge_loss, matthews_corrcoef, \
-                mean_squared_error, average_precision_score, auc_score
+from .metrics import (confusion_matrix, roc_curve, auc, precision_score,
+                      recall_score, fbeta_score, f1_score, zero_one_score,
+                      precision_recall_fscore_support, classification_report,
+                      precision_recall_curve, explained_variance_score,
+                      r2_score, zero_one, hinge_loss, matthews_corrcoef,
+                      mean_squared_error, average_precision_score, auc_score)
 
 from . import cluster
 from .cluster import adjusted_rand_score
diff --git a/sklearn/metrics/metrics.py b/sklearn/metrics/metrics.py
@@ -1027,30 +1027,6 @@ def mean_squared_error(y_true, y_pred):
     return np.mean((y_pred - y_true) ** 2)
 
 
-@deprecated("""Incorrectly returns the cumulated error: use mean_squared_error
-            instead; to be removed in v0.13""")
-def mean_square_error(y_true, y_pred):
-    """Cumulated square error regression loss
-
-    Positive floating point value: the best value is 0.0.
-
-    return the mean square error
-
-    Parameters
-    ----------
-    y_true : array-like
-
-    y_pred : array-like
-
-    Returns
-    -------
-    loss : float
-
-    """
-    y_true, y_pred = check_arrays(y_true, y_pred)
-    return np.linalg.norm(y_pred - y_true) ** 2
-
-
 def hinge_loss(y_true, pred_decision, pos_label=1, neg_label=-1):
     """
     Cumulated hinge loss (non-regularized).
diff --git a/sklearn/mixture/gmm.py b/sklearn/mixture/gmm.py
@@ -308,31 +308,6 @@ def eval(self, X):
         responsibilities = np.exp(lpr - logprob[:, np.newaxis])
         return logprob, responsibilities
 
-    @deprecated("""will be removed in v0.13;
-    use the score or predict method instead, depending on the question""")
-    def decode(self, X):
-        """Find most likely mixture components for each point in X.
-
-        DEPRECATED IN VERSION 0.11; WILL BE REMOVED IN VERSION 0.13.
-        use the score or predict method instead, depending on the question.
-
-        Parameters
-        ----------
-        X : array_like, shape (n, n_features)
-            List of n_features-dimensional data points.  Each row
-            corresponds to a single data point.
-
-        Returns
-        -------
-        logprobs : array_like, shape (n_samples,)
-            Log probability of each point in `obs` under the model.
-
-        components : array_like, shape (n_samples,)
-            Index of the most likelihod mixture components for each observation
-        """
-        logprob, posteriors = self.eval(X)
-        return logprob, posteriors.argmax(axis=1)
-
     def score(self, X):
         """Compute the log probability under the model.
 
@@ -381,16 +356,6 @@ def predict_proba(self, X):
         logprob, responsibilities = self.eval(X)
         return responsibilities
 
-    @deprecated("""will be removed in v0.13;
-    use the score or predict method instead, depending on the question""")
-    def rvs(self, n_samples=1, random_state=None):
-        """Generate random samples from the model.
-
-        DEPRECATED IN VERSION 0.11; WILL BE REMOVED IN VERSION 0.12
-        use sample instead
-        """
-        return self.sample(n_samples, random_state)
-
     def sample(self, n_samples=1, random_state=None):
         """Generate random samples from the model.
 
@@ -431,7 +396,7 @@ def sample(self, n_samples=1, random_state=None):
                     num_comp_in_X, random_state=random_state).T
         return X
 
-    def fit(self, X, **kwargs):
+    def fit(self, X):
         """Estimate model parameters with the expectation-maximization
         algorithm.
 
@@ -455,24 +420,6 @@ def fit(self, X, **kwargs):
             raise ValueError(
                 'GMM estimation with %s components, but got only %s samples' %
                 (self.n_components, X.shape[0]))
-        if kwargs:
-            warnings.warn("Setting parameters in the 'fit' method is"
-                          "deprecated and will be removed in 0.13. Set it on "
-                          "initialization instead.", DeprecationWarning,
-                          stacklevel=2)
-            # initialisations for in case the user still adds parameters to fit
-            # so things don't break
-            if 'n_iter' in kwargs:
-                self.n_iter = kwargs['n_iter']
-            if 'n_init' in kwargs:
-                if kwargs['n_init'] < 1:
-                    raise ValueError('GMM estimation requires n_init > 0.')
-                else:
-                    self.n_init = kwargs['n_init']
-            if 'params' in kwargs:
-                self.params = kwargs['params']
-            if 'init_params' in kwargs:
-                self.init_params = kwargs['init_params']
 
         max_log_prob = -np.infty
 
diff --git a/sklearn/svm/base.py b/sklearn/svm/base.py
@@ -113,7 +113,7 @@ def _pairwise(self):
         kernel = self.kernel
         return kernel == "precomputed" or hasattr(kernel, "__call__")
 
-    def fit(self, X, y, class_weight=None, sample_weight=None):
+    def fit(self, X, y, sample_weight=None):
         """Fit the SVM model according to the given training data.
 
         Parameters
@@ -160,13 +160,6 @@ def fit(self, X, y, class_weight=None, sample_weight=None):
             raise ValueError("The number of classes has to be greater than"
                     " one.")
 
-        if class_weight != None:
-            warnings.warn("'class_weight' is now an initialization parameter."
-                          "Using it in the 'fit' method is deprecated and "
-                          "will be removed in 0.13.", DeprecationWarning,
-                          stacklevel=2)
-            self.class_weight = class_weight
-
         sample_weight = np.asarray([] if sample_weight is None
                                       else sample_weight, dtype=np.float64)
         solver_type = LIBSVM_IMPL.index(self.impl)
@@ -624,7 +617,7 @@ def _get_solver_type(self):
                              + error_string)
         return self._solver_type_dict[solver_type]
 
-    def fit(self, X, y, class_weight=None):
+    def fit(self, X, y):
         """Fit the model according to the given training data.
 
         Parameters
@@ -651,13 +644,6 @@ def fit(self, X, y, class_weight=None):
             raise ValueError("The number of classes has to be greater than"
                     " one.")
 
-        if class_weight != None:
-            warnings.warn("'class_weight' is now an initialization parameter."
-                          "Using it in the 'fit' method is deprecated and "
-                          "will be removed in 0.13.", DeprecationWarning,
-                          stacklevel=2)
-            self.class_weight = class_weight
-
         X = atleast2d_or_csr(X, dtype=np.float64, order="C")
         y = np.asarray(y, dtype=np.float64).ravel()
 
