OwLouvain: Add data normalization option to PCA preprocessing

Orange/widgets/unsupervised/owlouvainclustering.py

@@ -17,6 +17,7 @@
 from Orange.clustering.louvain import table_to_knn_graph, Louvain
 from Orange.data import Table, DiscreteVariable
 from Orange.data.util import get_unique_names
+from Orange import preprocess
 from Orange.projection import PCA
 from Orange.widgets import widget, gui, report
 from Orange.widgets.settings import DomainContextHandler, ContextSetting, \
@@ -66,6 +67,7 @@ class Outputs:
 
     apply_pca = ContextSetting(True)
     pca_components = ContextSetting(_DEFAULT_PCA_COMPONENTS)
+    pca_normalize = ContextSetting(True)
     metric_idx = ContextSetting(0)
     k_neighbors = ContextSetting(_DEFAULT_K_NEIGHBORS)
     resolution = ContextSetting(1.)
@@ -106,11 +108,18 @@ def __init__(self):
             pca_box, self, "apply_pca", label="Apply PCA preprocessing",
             callback=self._invalidate_graph,
         )  # type: QCheckBox
+        pca_params_box = gui.vBox(pca_box)
+        # Make the difference between "Apply PCA" and PCA specific parameters clear
+        pca_params_box.layout().setContentsMargins(10, 0, 0, 0)
         self.pca_components_slider = gui.hSlider(
-            pca_box, self, "pca_components", label="Components: ", minValue=2,
+            pca_params_box, self, "pca_components", label="Components: ", minValue=2,
             maxValue=_MAX_PCA_COMPONENTS,
             callback=self._invalidate_pca_projection, tracking=False
         )  # type: QSlider
+        self.pca_normalize_cbx = gui.checkBox(
+            pca_params_box, self, "pca_normalize", label="Normalize data",
+            callback=self._invalidate_pca_projection,
+        )  # type: QCheckBox
 
         graph_box = gui.vBox(self.controlArea, "Graph parameters")
         self.metric_combo = gui.comboBox(
@@ -243,8 +252,8 @@ def commit(self):
         if graph is None:
             task = partial(
                 run_on_data, data, pca_components=pca_components,
-                k_neighbors=k_neighbors, metric=metric,
-                resolution=self.resolution, state=state
+                pca_normalize=self.pca_normalize, k_neighbors=k_neighbors,
+                metric=metric, resolution=self.resolution, state=state,
             )
         else:
             task = partial(
@@ -414,6 +423,17 @@ def set_data(self, data):
 
         self.info_label.setText("Clustering not yet run.")
 
+        self.pca_normalize_cbx.setDisabled(data.is_sparse())
+        # PCA doesn't support normalization on sparse data, as this would
+        # require centering and normalizing the matrix
+        if data.is_sparse():
+            self.pca_normalize_cbx.setToolTip(
+                "Data normalization is not supported on sparse matrices."
+            )
+            self.pca_normalize = False
+        else:
+            self.pca_normalize_cbx.setToolTip("")
+
         self.commit()
 
     def clear(self):
@@ -527,8 +547,10 @@ class Results(namespace):
     partition = None         # type: Optional[np.ndarray]
 
 
-def run_on_data(data, pca_components, k_neighbors, metric, resolution, state):
-    # type: (Table, Optional[int], int, str, float, TaskState) -> Results
+def run_on_data(
+        data, pca_components, pca_normalize, k_neighbors, metric, resolution, state
+):
+    # type: (Table, Optional[int], int, str, float, bool, TaskState) -> Results
     """
     Run the louvain clustering on `data`.
 
@@ -542,6 +564,8 @@ def run_on_data(data, pca_components, k_neighbors, metric, resolution, state):
     pca_components : Optional[int]
         If not `None` then the data is first projected onto first
         `pca_components` principal components.
+    pca_normalize : bool
+        If `True`, the data is first normalized before computing PCA.
     k_neighbors : int
         Passed to `table_to_knn_graph`
     metric : str
@@ -566,6 +590,9 @@ def run_on_data(data, pca_components, k_neighbors, metric, resolution, state):
         steps = 3
         state.set_status("Computing PCA...")
         pca = PCA(n_components=pca_components, random_state=0)
+        if pca_normalize:
+            pca.preprocessors += (preprocess.Normalize(),)
+
         data = res.pca_projection = pca(data)(data)
         assert isinstance(data, Table)
         state.set_partial_results(("pca_projection", res.pca_projection))

Orange/widgets/unsupervised/tests/test_owlouvain.py

@@ -3,6 +3,7 @@
 import numpy as np
 
 from Orange.data import Table, Domain, ContinuousVariable
+from Orange.preprocess import Normalize
 from Orange.widgets.tests.base import WidgetTest
 from Orange.widgets.unsupervised.owlouvainclustering import OWLouvainClustering
 
@@ -178,3 +179,32 @@ def test_deterministic_clustering(self):
 
         # Ensure that clustering was the same in both instances
         np.testing.assert_equal(result1.metas, result2.metas)
+
+    def test_normalize_data(self):
+        # Normalization should be checked by default
+        self.assertTrue(self.widget.controls.pca_normalize.isChecked())
+        with patch("Orange.preprocess.Normalize", wraps=Normalize) as normalize:
+            self.send_signal(self.widget.Inputs.data, self.iris)
+            self.wait_until_stop_blocking()
+            self.assertTrue(self.widget.controls.pca_normalize.isEnabled())
+            normalize.assert_called_once()
+
+        # Disable checkbox
+        self.widget.controls.pca_normalize.setChecked(False)
+        self.assertFalse(self.widget.controls.pca_normalize.isChecked())
+        with patch("Orange.preprocess.Normalize", wraps=Normalize) as normalize:
+            self.send_signal(self.widget.Inputs.data, self.iris)
+            self.wait_until_stop_blocking()
+            self.assertTrue(self.widget.controls.pca_normalize.isEnabled())
+            normalize.assert_not_called()
+
+        # Normalization shouldn't work on sparse data
+        self.widget.controls.pca_normalize.setChecked(True)
+        self.assertTrue(self.widget.controls.pca_normalize.isChecked())
+
+        sparse_data = self.iris.to_sparse()
+        with patch("Orange.preprocess.Normalize", wraps=Normalize) as normalize:
+            self.send_signal(self.widget.Inputs.data, sparse_data)
+            self.wait_until_stop_blocking()
+            self.assertFalse(self.widget.controls.pca_normalize.isEnabled())
+            normalize.assert_not_called()