Merge pull request #52 from jerneju/sparse-1

[FIX] Polynomial Classification: work on sparse

orangecontrib/educational/widgets/owpolynomialclassification.py

@@ -4,11 +4,12 @@
 import copy
 
 import numpy as np
+import scipy.sparse as sp
+from scipy.interpolate import splprep, splev
 from scipy.ndimage.filters import gaussian_filter
+
 from AnyQt.QtCore import Qt
 from AnyQt.QtGui import QPixmap, QColor, QIcon
-from AnyQt.QtWidgets import QSizePolicy
-from scipy.interpolate import splprep, splev
 
 from Orange.base import Learner as InputLearner
 from Orange.data import (
@@ -531,8 +532,9 @@ def select_data(self):
         cols = []
         for attr in (attr_x, attr_y):
             subset = self.data[:, attr]
-            cols.append(subset.X)
+            cols.append(subset.X if not sp.issparse(subset.X) else subset.X.toarray())
         x = np.column_stack(cols)
+        y_c = self.data.Y[:, None] if not sp.issparse(self.data.Y) else self.data.Y.toarray()
 
         if np.isnan(x).all(axis=0).any():
             self.Error.all_none_data()
@@ -554,7 +556,7 @@ def select_data(self):
         y = [(0 if d.get_class().value == self.target_class else 1)
              for d in self.data]
 
-        return Table(domain, x, y, self.data.Y[:, None])
+        return Table(domain, x, y, y_c)
 
     def apply(self):
         """
@@ -563,7 +565,7 @@ def apply(self):
         self.send_learner()
         self.update_model()
         self.send_coefficients()
-        if None in (self.data, self.model):
+        if any(a is None for a in (self.data, self.model)):
             self.set_empty_plot()
         else:
             self.replot()

orangecontrib/educational/widgets/tests/test_owgradientdescent.py

@@ -983,3 +983,7 @@ def send_sparse_data(data):
             class_vars=data.domain.attributes[3:] + data.domain.class_vars
         )
         send_sparse_data(data.transform(domain))
+
+
+if __name__ == "__main__":
+    unittest.main()

orangecontrib/educational/widgets/tests/test_owkmeans.py

@@ -335,3 +335,8 @@ def test_data_nan_column(self):
         data = data.transform(domain)
         data[:, 0] = np.nan
         self.send_signal(self.widget.Inputs.data, data)
+
+
+if __name__ == "__main__":
+    import unittest
+    unittest.main()

orangecontrib/educational/widgets/tests/test_owpolynomialclassification.py

@@ -1,9 +1,10 @@
 from functools import reduce
 import unittest
 
-from Orange.regression import LinearRegressionLearner
+import scipy.sparse as sp
 from numpy.testing import assert_array_equal
 
+from Orange.regression import LinearRegressionLearner
 from Orange.widgets.tests.base import WidgetTest
 from Orange.data import Table, ContinuousVariable, Domain, DiscreteVariable
 from Orange.classification import (
@@ -606,3 +607,31 @@ def test_no_data_contour(self):
         GH-50
         """
         self.widget.contours_enabled_checkbox.click()
+
+    def test_sparse(self):
+        """
+        Do not crash on sparse data. Convert used
+        sparse columns to numpy array.
+        GH-52
+        """
+        w = self.widget
+
+        def send_sparse_data(data):
+            data.X = sp.csr_matrix(data.X)
+            data.Y = sp.csr_matrix(data.Y)
+            self.send_signal(w.Inputs.data, data)
+
+        # one class variable
+        send_sparse_data(Table("iris"))
+
+        # two class variables
+        data = Table("iris")
+        domain = Domain(
+            attributes=data.domain.attributes[:3],
+            class_vars=data.domain.attributes[3:] + data.domain.class_vars
+        )
+        send_sparse_data(data.transform(domain))
+
+
+if __name__ == "__main__":
+    unittest.main()

orangecontrib/educational/widgets/tests/test_owpolynomialregression.py

@@ -250,3 +250,8 @@ def test_data_nan_row(self):
         data = Table("iris")[::50]
         data.X[0] = np.nan
         self.send_signal(self.widget.Inputs.data, data)
+
+
+if __name__ == "__main__":
+    import unittest
+    unittest.main()