Adding default_parameters classproperty to components and pipelines. (#…

…879)

* Adding default_parameters classproperty to components and pipelines.

* Adding PR 879 to changelog

* Updating test_default_parameters so that classes that cannot be imported are skipped.

* Fixing typo in cannot_check_because_base_or_not_installed

* Fixing documentation and splitting test_default_parameters.py into test_pipelines.py and test_components.py

* Adding DateTimeFeaturizer to test_default_parameters.

* Adding default_parameters classproperty to components and pipelines.

* Adding PR 879 to changelog

* Updating test_default_parameters so that classes that cannot be imported are skipped.

* Fixing typo in cannot_check_because_base_or_not_installed

* Fixing documentation and splitting test_default_parameters.py into test_pipelines.py and test_components.py

* Adding DateTimeFeaturizer to test_default_parameters.

* Fixing imports in test_components

* Fixing redundant import in test_components.py

* Adding default_parameters to estimator_class, pipeline_class, transformer_class rst files.

* Modifying implementation of default_properties so that DateTimeFeaturization does not have custom code.

docs/source/_templates/estimator_class.rst

@@ -5,14 +5,16 @@
 .. inheritance-diagram:: {{ objname }}
 
 .. autoclass:: {{ objname }}
-   {% set class_attributes = ['name', 'model_family', 'hyperparameter_ranges', 'supported_problem_types'] %}
+   {% set class_attributes = ['name', 'model_family', 'hyperparameter_ranges', 'supported_problem_types',
+                              'default_parameters'] %}
 
    {% block attributes %}
    .. Class attributes:
    .. autoattribute:: name
    .. autoattribute:: model_family
    .. autoattribute:: supported_problem_types
    .. autoattribute:: hyperparameter_ranges
+   .. autoattribute:: default_parameters
    {% endblock %}
 
    {% block instance_attributes %}

docs/source/_templates/pipeline_class.rst

@@ -5,7 +5,9 @@
 .. inheritance-diagram:: {{ objname }}
 
 .. autoclass:: {{ objname }}
-   {% set class_attributes = ['name', 'custom_name', 'summary', 'component_graph', 'problem_type', 'model_family', 'hyperparameters', 'custom_hyperparameters'] %}
+   {% set class_attributes = ['name', 'custom_name', 'summary', 'component_graph', 'problem_type',
+                              'model_family', 'hyperparameters', 'custom_hyperparameters',
+                              'default_parameters'] %}
 
 
    {% block attributes %}
@@ -19,6 +21,7 @@
    .. autoattribute:: model_family
    .. autoattribute:: hyperparameters
    .. autoattribute:: custom_hyperparameters
+   .. autoattribute:: default_parameters
    {% endblock %}
 
    {% block instance_attributes %}

docs/source/_templates/transformer_class.rst

@@ -6,13 +6,14 @@
 
 .. autoclass:: {{ objname }}
 
-   {% set class_attributes = ['name', 'model_family', 'hyperparameter_ranges'] %}
+   {% set class_attributes = ['name', 'model_family', 'hyperparameter_ranges', 'default_parameters'] %}
 
    {% block attributes %}
    .. Class attributes:
    .. autoattribute:: name
    .. autoattribute:: model_family
    .. autoattribute:: hyperparameter_ranges
+   .. autoattribute:: default_parameters
    {% endblock %}
 
    {% block instance_attributes %}

docs/source/changelog.rst

@@ -18,6 +18,7 @@ Changelog
         * Added new utility functions necessary for generating dynamic preprocessing pipelines :pr:`852`
         * Added kwargs to all components :pr:`863`
         * Added SelectColumns transformer :pr:`873`
+        * Added `default_parameters` class property to components and pipelines :pr:`879`
     * Fixes
         * Fixed bug where SimpleImputer cannot handle dropped columns :pr:`846`
         * Fixed bug where PerColumnImputer cannot handle dropped columns :pr:`855`

evalml/pipelines/components/component_base.py

@@ -2,13 +2,19 @@
 from abc import ABC, abstractmethod
 
 from evalml.exceptions import MethodPropertyNotFoundError
-from evalml.utils import get_logger, get_random_state, log_subtitle
+from evalml.utils import (
+    classproperty,
+    get_logger,
+    get_random_state,
+    log_subtitle
+)
 
 logger = get_logger(__file__)
 
 
 class ComponentBase(ABC):
     """Base class for all components."""
+    _default_parameters = None
 
     def __init__(self, parameters=None, component_obj=None, random_state=0, **kwargs):
         self.random_state = get_random_state(random_state)
@@ -32,6 +38,21 @@ def parameters(self):
         """Returns the parameters which were used to initialize the component"""
         return copy.copy(self._parameters)
 
+    @classproperty
+    def default_parameters(cls):
+        """Returns the default parameters for this component.
+
+         Our convention is that Component.default_parameters == Component().parameters.
+
+         Returns:
+             dict: default parameters for this component.
+        """
+
+        if cls._default_parameters is None:
+            cls._default_parameters = cls().parameters
+
+        return cls._default_parameters
+
     def clone(self, random_state=0):
         """Constructs a new component with the same parameters
 

evalml/pipelines/components/utils.py

@@ -26,6 +26,7 @@
     PerColumnImputer,
     RFClassifierSelectFromModel,
     RFRegressorSelectFromModel,
+    SelectColumns,
     SimpleImputer,
     StandardScaler
 )

evalml/pipelines/pipeline_base.py

@@ -270,6 +270,20 @@ def parameters(self):
         """
         return {c.name: copy.copy(c.parameters) for c in self.component_graph if c.parameters}
 
+    @classproperty
+    def default_parameters(cls):
+        """Returns the default parameter dictionary for this pipeline.
+
+        Returns:
+            dict: dictionary of all component default parameters.
+        """
+        defaults = {}
+        for c in cls.component_graph:
+            component = handle_component_class(c)
+            if component.default_parameters:
+                defaults[component.name] = component.default_parameters
+        return defaults
+
     @property
     def feature_importance(self):
         """Return importance associated with each feature. Features dropped by feature selection are excluded"""

evalml/tests/component_tests/test_components.py

@@ -10,6 +10,7 @@
 from evalml.model_family import ModelFamily
 from evalml.pipelines.components import (
     ComponentBase,
+    DateTimeFeaturization,
     DropColumns,
     ElasticNetClassifier,
     ElasticNetRegressor,
@@ -23,6 +24,7 @@
     RandomForestClassifier,
     RandomForestRegressor,
     RFClassifierSelectFromModel,
+    SelectColumns,
     SimpleImputer,
     StandardScaler,
     Transformer,
@@ -441,16 +443,29 @@ def test_transformer_transform_output_type(X_y):
                   .format(component_class.name, type(X),
                           X.columns if isinstance(X, pd.DataFrame) else None, type(y),
                           y.name if isinstance(y, pd.Series) else None))
+
             component = component_class()
+
             component.fit(X, y=y)
             transform_output = component.transform(X, y=y)
             assert isinstance(transform_output, pd.DataFrame)
-            assert transform_output.shape == X.shape
-            assert (transform_output.columns == X_cols_expected).all()
+
+            if isinstance(component, SelectColumns):
+                assert transform_output.shape == (X.shape[0], 0)
+                assert isinstance(transform_output.columns, pd.Index)
+            else:
+                assert transform_output.shape == X.shape
+                assert (transform_output.columns == X_cols_expected).all()
+
             transform_output = component.fit_transform(X, y=y)
             assert isinstance(transform_output, pd.DataFrame)
-            assert transform_output.shape == X.shape
-            assert (transform_output.columns == X_cols_expected).all()
+
+            if isinstance(component, SelectColumns):
+                assert transform_output.shape == (X.shape[0], 0)
+                assert isinstance(transform_output.columns, pd.Index)
+            else:
+                assert transform_output.shape == X.shape
+                assert (transform_output.columns == X_cols_expected).all()
 
 
 def test_estimator_predict_output_type(X_y):
@@ -494,3 +509,12 @@ def test_estimator_predict_output_type(X_y):
             assert isinstance(predict_proba_output, pd.DataFrame)
             assert predict_proba_output.shape == (len(y), len(np.unique(y)))
             assert (predict_proba_output.columns == y_cols_expected).all()
+
+
+components = list(all_components().items()) + [(DateTimeFeaturization.name, DateTimeFeaturization)]
+
+
+@pytest.mark.parametrize("class_name,cls", components)
+def test_default_parameters(class_name, cls):
+
+    assert cls.default_parameters == cls().parameters, f"{class_name}'s default parameters don't match __init__."

evalml/tests/component_tests/test_utils.py

@@ -14,9 +14,9 @@
 
 def test_all_components(has_minimal_dependencies):
     if has_minimal_dependencies:
-        assert len(all_components()) == 17
+        assert len(all_components()) == 18
     else:
-        assert len(all_components()) == 21
+        assert len(all_components()) == 22
 
 
 def make_mock_import_module(libs_to_exclude):
@@ -29,7 +29,7 @@ def _import_module(library):
 
 @patch('importlib.import_module', make_mock_import_module({'xgboost', 'catboost'}))
 def test_all_components_core_dependencies_mock():
-    assert len(all_components()) == 17
+    assert len(all_components()) == 18
 
 
 def test_handle_component_class_names():

evalml/tests/pipeline_tests/test_pipelines.py

@@ -866,3 +866,9 @@ def test_clone_fitted(X_y, lr_pipeline):
     pipeline_clone.fit(X, y)
     X_t_clone = pipeline_clone.predict_proba(X)
     pd.testing.assert_frame_equal(X_t, X_t_clone)
+
+
+@pytest.mark.parametrize("cls", all_pipelines())
+def test_pipeline_default_parameters(cls):
+
+    assert cls.default_parameters == cls({}).parameters, f"{cls.__name__}'s default parameters don't match __init__."