Partial dependence plots

alibi/explainers/__init__.py

@@ -9,7 +9,7 @@
 from alibi.explainers.anchors.anchor_image import AnchorImage
 from alibi.explainers.cfrl_base import CounterfactualRL
 from alibi.explainers.cfrl_tabular import CounterfactualRLTabular
-from alibi.explainers.partial_dependence import PartialDependence, plot_pd
+from alibi.explainers.partial_dependence import PartialDependence, TreePartialDependence, plot_pd
 from alibi.explainers.similarity.grad import GradientSimilarity
 
 
@@ -50,7 +50,7 @@
     "CounterfactualRLTabular",
     "plot_ale",
     "PartialDependence",
-    "PDEstimatorWrapper",
+    "TreePartialDependence",
     "plot_pd",
     "IntegratedGradients",
     "KernelShap",

alibi/explainers/tests/conftest.py

@@ -7,7 +7,6 @@
 
 from sklearn.linear_model import LogisticRegression, LinearRegression
 from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
-from sklearn.svm import SVC
 
 from alibi.explainers import ALE
 from alibi.explainers import AnchorTabular
@@ -90,37 +89,6 @@ def movie_sentiment_data():
 #  ble to parametrize with module name and param dict to have only 1
 #  such fixture
 
-@pytest.fixture(scope='module')
-def svc_classifier(request):
-    """
-    Trains a support vector classifier. Because it is scoped
-    at module level, the state of this  fixture should not be
-    mutated during test - if you need to do so, please copy the
-    objects returned.
-    """
-
-    is_preprocessor = False
-    preprocessor = None
-    # this fixture should be parametrised with a fixture that
-    # returns a dataset dictionary with specified attributes
-    # see test_anchor_tabular for a usage example
-    data = request.param
-
-    if data['preprocessor']:
-        is_preprocessor = True
-        preprocessor = data['preprocessor']
-
-    np.random.seed(0)
-    clf = SVC()
-
-    if is_preprocessor:
-        clf.fit(preprocessor.transform(data['X_train']), data['y_train'])
-    else:
-        clf.fit(data['X_train'], data['y_train'])
-
-    return clf, preprocessor
-
-
 @pytest.fixture(scope='module')
 def rf_classifier(request):
     """