Merge pull request #2180 from VesnaT/manifold_sparse

OWManifoldLearning: Show error msg when data is sparse
biolab · Apr 7, 2017 · 68727b4 · 68727b4
2 parents d514b5b + c27cce1
commit 68727b4
Show file tree

Hide file tree

Showing 2 changed files with 37 additions and 21 deletions.
diff --git a/Orange/widgets/unsupervised/owmanifoldlearning.py b/Orange/widgets/unsupervised/owmanifoldlearning.py
@@ -179,6 +179,7 @@ class OWManifoldLearning(OWWidget):
     class Error(OWWidget.Error):
         n_neighbors_too_small = Msg("Neighbors must be greater than {}.")
         manifold_error = Msg("{}")
+        sparse_not_supported = Msg("Sparse data is not supported.")
 
     def __init__(self):
         self.data = None
@@ -235,27 +236,30 @@ def apply(self):
         data = None
         self.clear_messages()
         if self.data:
-            with self.progressBar():
-                self.progressBarSet(10)
-                domain = Domain([ContinuousVariable("C{}".format(i))
-                                 for i in range(self.n_components)],
-                                self.data.domain.class_vars,
-                                self.data.domain.metas)
-
-                method = self.MANIFOLD_METHODS[self.manifold_method_index]
-                projector = method(**self.get_method_parameters())
-                try:
-                    self.progressBarSet(20)
-                    X = projector(self.data).embedding_
-                    data = Table(domain, X, self.data.Y, self.data.metas)
-                except ValueError as e:
-                    if e.args[0] == "for method='hessian', n_neighbors must " \
-                                    "be greater than [n_components * (" \
-                                    "n_components + 3) / 2]":
-                        n = self.n_components * (self.n_components + 3) / 2
-                        self.Error.n_neighbors_too_small("{}".format(n))
-                    else:
-                        self.Error.manifold_error(e.args[0])
+            if self.data.is_sparse():
+                self.Error.sparse_not_supported()
+            else:
+                with self.progressBar():
+                    self.progressBarSet(10)
+                    domain = Domain([ContinuousVariable("C{}".format(i))
+                                     for i in range(self.n_components)],
+                                    self.data.domain.class_vars,
+                                    self.data.domain.metas)
+
+                    method = self.MANIFOLD_METHODS[self.manifold_method_index]
+                    projector = method(**self.get_method_parameters())
+                    try:
+                        self.progressBarSet(20)
+                        X = projector(self.data).embedding_
+                        data = Table(domain, X, self.data.Y, self.data.metas)
+                    except ValueError as e:
+                        if e.args[0] == "for method='hessian', n_neighbors " \
+                                        "must be greater than [n_components" \
+                                        " * (n_components + 3) / 2]":
+                            n = self.n_components * (self.n_components + 3) / 2
+                            self.Error.n_neighbors_too_small("{}".format(n))
+                        else:
+                            self.Error.manifold_error(e.args[0])
         self.send("Transformed data", data)
 
     def get_method_parameters(self):

diff --git a/Orange/widgets/unsupervised/tests/test_owmanifoldlearning.py b/Orange/widgets/unsupervised/tests/test_owmanifoldlearning.py
@@ -1,6 +1,7 @@
 # Test methods with long descriptive names can omit docstrings
 # pylint: disable=missing-docstring
 import numpy as np
+from scipy import sparse
 
 from Orange.data import Table
 from Orange.widgets.unsupervised.owmanifoldlearning import OWManifoldLearning
@@ -61,3 +62,14 @@ def _compare_tables(self, _output, n_components):
         self.assertEqual((len(self.iris), n_components), _output.X.shape)
         np.testing.assert_array_equal(self.iris.Y, _output.Y)
         np.testing.assert_array_equal(self.iris.metas, _output.metas)
+
+    def test_sparse_data(self):
+        data = Table("iris")
+        data.X = sparse.csr_matrix(data.X)
+        self.assertTrue(sparse.issparse(data.X))
+        self.send_signal("Data", data)
+        self.widget.apply_button.button.click()
+        self.assertTrue(self.widget.Error.sparse_not_supported.is_shown())
+        self.send_signal("Data", None)
+        self.widget.apply_button.button.click()
+        self.assertFalse(self.widget.Error.sparse_not_supported.is_shown())