Remove append original (#29)

* remove all traces of append_original

* remove all traces of append_original from UI

* add identity transformer

* update docs for removing append_original

docs/walkthrough.rst

@@ -406,7 +406,10 @@ To set the hyperparameters of the secondary learner, enter the following into th
 
 This should keep the Logistic Regression at its default values. If you'll notice, the format required for this code block is exactly the same as that required when creating a single base learner.
 
-There's an additional checkbox you can tick to append the original features to the base learners' meta-features. Leave it unchecked for now, and go ahead and **Create new ensemble**.
+.. admonition:: Appending original features to the meta-features
+
+   As of 0.4.0, Xcessiv no longer provides the checkbox option to append the original features to the meta-features. However, this functionality can be achieved by using the preset "Identity Transformer" and including this as a base learner in the ensemble. This approach reinforces the fact that a base learner is simply something that generates meta-features from features, and one that lets all features pass through unchanged fits this description.
+
 
 After a short time, your ensemble should finish processing, and you'll be able to see its performance. Here we get an accuracy of 0.968, which is higher than any individual base learner.
 

xcessiv/models.py

@@ -367,17 +367,15 @@ class StackedEnsemble(Base):
     base_learner_origin = relationship('BaseLearnerOrigin', back_populates='stacked_ensembles')
     secondary_learner_hyperparameters = Column(JsonEncodedDict)
     individual_score = Column(JsonEncodedDict)
-    append_original = Column(Boolean)
     job_status = Column(Text)
     job_id = Column(Text)
     description = Column(JsonEncodedDict)
 
     def __init__(self, secondary_learner_hyperparameters, base_learners,
-                 base_learner_origin, append_original, job_status):
+                 base_learner_origin, job_status):
         self.base_learner_origin = base_learner_origin
         self.secondary_learner_hyperparameters = secondary_learner_hyperparameters
         self.base_learners = base_learners
-        self.append_original = append_original
         self.individual_score = dict()
         self.job_status = job_status
         self.job_id = None
@@ -461,12 +459,10 @@ def export_as_code(self, cv_source):
                              '(base_learners=base_learner_list_{},' \
                              ' meta_feature_generators=meta_feature_generators_list_{},' \
                              ' secondary_learner=secondary_learner_{},' \
-                             ' cv_function=return_splits_iterable,' \
-                             ' append_original={})\n'.format(
+                             ' cv_function=return_splits_iterable)\n'.format(
             rand_value,
             rand_value,
-            rand_value,
-            self.append_original
+            rand_value
         )
 
         return base_learner_code
@@ -571,8 +567,7 @@ def export_as_package(self, package_path, cv_source):
         builder_source += '\nxcessiv_ensemble = XcessivStackedEnsemble(base_learners=base_learners,' \
                           ' meta_feature_generators=meta_feature_generators,' \
                           ' secondary_learner=metalearner.base_learner,' \
-                          ' cv_function=return_splits_iterable,' \
-                          ' append_original={})\n'.format(self.append_original)
+                          ' cv_function=return_splits_iterable)\n'
 
         with open(os.path.join(package_path, 'builder.py'), 'wb') as f:
             f.write(builder_source.encode('utf8'))
@@ -588,6 +583,5 @@ def serialize(self):
             description=self.description,
             base_learner_origin_id=self.base_learner_origin_id,
             base_learner_ids=list(map(lambda x: x.id, self.base_learners)),
-            number_of_base_learners=len(self.base_learners),
-            append_original=self.append_original
+            number_of_base_learners=len(self.base_learners)
         )

xcessiv/presets/learnersetting.py

@@ -27,7 +27,10 @@
     'sklearn_decision_tree_regressor',
     'sklearn_linear_regression',
     'sklearn_adaboost_regressor',
-    'xgboost_regressor'
+    'xgboost_regressor',
+
+    # Transformers
+    'identity_transformer'
 ]
 
 # Classifiers
@@ -157,3 +160,10 @@
     'source': learnersource.xgboost_regressor_source,
     'meta_feature_generator': 'predict'
 }
+
+
+identity_transformer = {
+    'name': 'Identity Transformer',
+    'source': learnersource.identity_transformer_source,
+    'meta_feature_generator': 'transform'
+}

xcessiv/presets/learnersource.py

@@ -26,7 +26,10 @@
     'sklearn_decision_tree_regressor_source',
     'sklearn_linear_regression_source',
     'sklearn_adaboost_regressor_source',
-    'xgboost_regressor_source'
+    'xgboost_regressor_source',
+
+    # Transformers
+    'identity_transformer_source'
 ]
 
 sklearn_random_forest_classifier_source = \
@@ -154,3 +157,9 @@
 
 base_learner = XGBRegressor(seed=8)
 """
+
+identity_transformer_source = \
+    """from sklearn.preprocessing import FunctionTransformer
+
+base_learner = FunctionTransformer()
+"""

xcessiv/rqtasks.py

@@ -437,9 +437,6 @@ def evaluate_stacked_ensemble(path, ensemble_id):
             indices = np.concatenate(indices_list)
             X, y = X[indices], y[indices]
 
-            if stacked_ensemble.append_original:
-                secondary_features = np.concatenate((secondary_features, X), axis=1)
-
             est = stacked_ensemble.return_secondary_learner()
 
             return_splits_iterable_stacked_ensemble = functions.import_object_from_string_code(

xcessiv/stacker.py

@@ -5,14 +5,13 @@
 class XcessivStackedEnsemble(_BasePipeline):
     """Contains the class for the Xcessiv stacked ensemble"""
     def __init__(self, base_learners, meta_feature_generators,
-                 secondary_learner, cv_function, append_original):
+                 secondary_learner, cv_function):
         super(XcessivStackedEnsemble, self).__init__()
 
         self.base_learners = base_learners
         self.meta_feature_generators = meta_feature_generators
         self.secondary_learner = secondary_learner
         self.cv_function = cv_function
-        self.append_original = append_original
         self._named_learners = [('bl{}'.format(idx), base_learner) for idx, base_learner
                                in enumerate(base_learners)]
         self._named_learners.append(('secondary-learner', secondary_learner))
@@ -67,12 +66,6 @@ def fit(self, X, y):
 
         print('Fitting meta-learner')
 
-        if self.append_original:
-            all_learner_meta_features = np.concatenate(
-                (all_learner_meta_features, X[test_indices]),
-                axis=1
-            )
-
         self.secondary_learner.fit(all_learner_meta_features, y[test_indices])
 
         return self
@@ -100,11 +93,6 @@ def _process_using_meta_feature_generator(self, X, meta_feature_generator):
             all_learner_meta_features.append(single_learner_meta_features)
 
         all_learner_meta_features = np.concatenate(all_learner_meta_features, axis=1)
-        if self.append_original:
-            all_learner_meta_features = np.concatenate(
-                (all_learner_meta_features, X),
-                axis=1
-            )
 
         out = getattr(self.secondary_learner, meta_feature_generator)(all_learner_meta_features)
 

xcessiv/ui/src/Ensemble/EnsembleBuilder.js

@@ -8,7 +8,7 @@ import Select from 'react-select'
 import CodeMirror from 'react-codemirror';
 import 'codemirror/lib/codemirror.css';
 import 'codemirror/mode/python/python';
-import { Checkbox, Button, Glyphicon } from 'react-bootstrap';
+import { Button, Glyphicon } from 'react-bootstrap';
 
 
 const defaultSourceParams = [
@@ -24,8 +24,7 @@ class EnsembleBuilder extends Component {
     super(props);
     this.state = {
       selectedValue: null,
-      source: defaultSourceParams,
-      appendOriginal: false
+      source: defaultSourceParams
     };
   }
 
@@ -66,19 +65,12 @@ class EnsembleBuilder extends Component {
           onChange={(src) => this.setState({source: src})} 
           options={options}
         />
-        <Checkbox
-          checked={this.state.appendOriginal} 
-          onChange={() => this.setState((prevState) => 
-            ({appendOriginal: !prevState.appendOriginal}))}
-        >
-          Append original features to secondary features
-        </Checkbox>
         <Button 
           block
           disabled={buttonDisabled}
           bsStyle='primary'
           onClick={() => this.props.createStackedEnsemble(
-            this.state.selectedValue.value, this.state.source, this.state.appendOriginal)}>
+            this.state.selectedValue.value, this.state.source)}>
           <Glyphicon glyph="plus" />
           {' Create new ensemble'}
         </Button>

xcessiv/ui/src/Ensemble/ListEnsemble.js

@@ -106,7 +106,6 @@ class ListEnsemble extends Component {
     const cols = [
       {label: 'ID', value: 'id', width: 50}, 
       {label: 'Secondary Learner ID', value: 'base_learner_origin_id', width: 100},
-      {label: 'Appended Original Features', value: 'append_original', width: 100},
       {label: 'Number of Base Learners', value: 'number_of_base_learners', width: 100}
     ]
     return cols.map((obj) => {

xcessiv/ui/src/containers/ContainerBaseLearner.js

@@ -530,12 +530,11 @@ class ContainerBaseLearner extends Component {
 
   // Create new stacked ensemble
   createStackedEnsemble(base_learner_ids, base_learner_origin_id, 
-    secondary_learner_hyperparameters_source, append_original) {
+    secondary_learner_hyperparameters_source) {
     var payload = {
       base_learner_ids, 
       base_learner_origin_id, 
-      secondary_learner_hyperparameters_source,
-      append_original
+      secondary_learner_hyperparameters_source
     };
 
     fetch(
@@ -715,8 +714,8 @@ class ContainerBaseLearner extends Component {
           optionsBaseLearnerOrigins={optionsBaseLearnerOrigins}
           checkedOptions={checkedOptions}
           setCheckedBaseLearners={(checkedArray) => this.setState({checkedBaseLearners: ImSet(checkedArray)})}
-          createStackedEnsemble={(bloId, hp, appendOriginal) => 
-            this.createStackedEnsemble(this.state.checkedBaseLearners, bloId, hp, appendOriginal)}
+          createStackedEnsemble={(bloId, hp) => 
+            this.createStackedEnsemble(this.state.checkedBaseLearners, bloId, hp)}
         />
         <ListEnsemble 
           path={this.props.path}

xcessiv/views.py

@@ -556,7 +556,6 @@ def create_new_stacked_ensemble():
                 secondary_learner_hyperparameters=hyperparameters,
                 base_learners=base_learners,
                 base_learner_origin=base_learner_origin,
-                append_original=req_body['append_original'],
                 job_status='queued'
             )