0.17

Author: atavory

.travis.yml

@@ -21,7 +21,7 @@ install:
   - sudo pip install python-coveralls
   - sudo pip install codecov
   - if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then
-      conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION six numpy scipy scikit-learn=0.18 pandas coverage jupyter seaborn xgboost tensorflow;
+      conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION six numpy scipy scikit-learn=0.17 pandas coverage jupyter seaborn xgboost tensorflow;
     else
       conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION six numpy scipy scikit-learn pandas coverage jupyter seaborn xgboost tensorflow;
     fi

docs/source/extending.rst

@@ -101,9 +101,12 @@ We can now use this as a regular step. If we fit it on ``df`` and transform it o
 
 We can, however, now use it for fitting on one ``DataFrame``, and transforming another:
 
-    >>> from sklearn import model_selection
+    >>> try:
+    ...     from sklearn.model_selection import train_test_split
+    ... except: # Older sklearn versions
+    ...     from ibex.sklearn.cross_val_predict import train_test_split
     >>>
-    >>> tr, te = model_selection.train_test_split(df, random_state=3)
+    >>> tr, te = train_test_split(df, random_state=3)
     >>> GroupbyAggregator('a').fit(tr).transform(te)
          b    c
     0  0...  2...

docs/source/feature_union.rst

@@ -64,11 +64,11 @@ The output using this, however, discards the meaning of the columns:
     >>> trn.fit_transform(iris[features], iris['class'])
               pca                 selectkbest
            comp_0    comp_1 petal length (cm)
-    0   -2.684207  0.326607               1.4
-    1   -2.715391 -0.169557               1.4
-    2   -2.889820 -0.137346               1.3
-    3   -2.746437 -0.311124               1.5
-    4   -2.728593  0.333925               1.4
+    0   -2.684207 ...0.326607               1.4
+    1   -2.715391 ...0.169557               1.4
+    2   -2.889820 ...0.137346               1.3
+    3   -2.746437 ...0.311124               1.5
+    4   -2.728593 ...0.333925               1.4
 	...
 
 A better way would be to combine this with :func:`ibex.trans`:
@@ -79,11 +79,11 @@ A better way would be to combine this with :func:`ibex.trans`:
     >>> trn.fit_transform(iris[features], iris['class'])
         functiontransformer_0           functiontransformer_1
                           pc1       pc2                  best
-    0               -2.684207  0.326607                   1.4
-    1               -2.715391 -0.169557                   1.4
-    2               -2.889820 -0.137346                   1.3
-    3               -2.746437 -0.311124                   1.5
-    4               -2.728593  0.333925                   1.4
+    0               -2.684207 ...0.326607                   1.4
+    1               -2.715391 ...0.169557                   1.4
+    2               -2.889820 ...0.137346                   1.3
+    3               -2.746437 ...0.311124                   1.5
+    4               -2.728593 ...0.333925                   1.4
 	...
 
 

examples/iris_feature_importance.ipynb

@@ -42,18 +42,16 @@ def _from_pickle(
         est,
         params,
         extra_methods,
-        extra_attribs,
-    post_op):
-    cls = frame_ex(est, extra_methods, extra_attribs, post_op)
+        extra_attribs):
+    cls = frame_ex(est, extra_methods, extra_attribs)
     est = cls(**params)
     return est
 
 
 def make_adapter(
         est,
         extra_methods,
-        extra_attribs,
-    post_op):
+        extra_attribs):
     from ._base import FrameMixin
 
     extra_attribs_d = {fn.__name__: fn for fn in extra_attribs}
@@ -322,15 +320,12 @@ def __getattribute__(self, name):
         def __reduce__(self):
             if not self.__module__.startswith('ibex'):
                 raise TypeError('Cannot serialize a subclass of this type; please use composition instead')
-            return (_from_pickle, (est, self.get_params(deep=True), extra_methods, extra_attribs, post_op))
-
-    if post_op is not None:
-        post_op(_Adapter)
+            return (_from_pickle, (est, self.get_params(deep=True), extra_methods, extra_attribs))
 
     return _Adapter
 
 
-def frame_ex(est, extra_methods=(), extra_attribs=(), post_op=None):
+def frame_ex(est, extra_methods=(), extra_attribs=()):
     from ._base import FrameMixin
 
     if isinstance(est, FrameMixin):
@@ -341,7 +336,7 @@ def frame_ex(est, extra_methods=(), extra_attribs=(), post_op=None):
         f = frame(type(est))(**params)
         return f
 
-    _Adapter = make_adapter(est, extra_methods, extra_attribs, post_op)
+    _Adapter = make_adapter(est, extra_methods, extra_attribs)
 
     update_class_wrapper(_Adapter, est)
 

ibex/_adapter.py

@@ -83,17 +83,20 @@ def cross_val_predict(
     Example:
 
         >>> import pandas as pd
-        >>> from ibex.sklearn import model_selection as pd_model_selection
-        >>> from ibex.sklearn import linear_model as pd_linear_model
+        >>> from ibex.sklearn.linear_model import LinearRegression
+        >>> try:
+        ...     from ibex.sklearn.model_selection import cross_val_predict
+        ... except: # Older sklearn versions
+        ...     from ibex.sklearn.cross_val_predict import cross_val_predict
 
         >>> df = pd.DataFrame({
         ...         'x': range(100),
         ...         'y': range(100),
         ...     },
         ...     index=['i%d' % i for i in range(100)])
 
-        >>> pd_model_selection.cross_val_predict(
-        ...     pd_linear_model.LinearRegression(),
+        >>> cross_val_predict(
+        ...     LinearRegression(),
         ...     df[['x']],
         ...     df['y'])
         i0     ...

ibex/sklearn/_cross_val_predict.py

@@ -41,11 +41,11 @@
 
         >>> PdPCA(n_components=2).fit(iris[features], iris['class']).transform(iris[features])
             comp_0    comp_1
-        0   -2.684207  0.326607
-        1   -2.715391 -0.169557
-        2   -2.889820 -0.137346
-        3   -2.746437 -0.311124
-        4   -2.728593  0.333925
+        0   -2.684207 ...0.326607
+        1   -2.715391 ...0.169557
+        2   -2.889820 ...0.137346
+        3   -2.746437 ...0.311124
+        4   -2.728593 ...0.333925
         ...
 
 """