[ENH] OwLouvain: Add normalize data checkbox to PCA preprocessing

Orange/preprocess/normalize.py

@@ -11,21 +11,24 @@ class Normalizer(Reprable):
     def __init__(self,
                  zero_based=True,
                  norm_type=Normalize.NormalizeBySD,
-                 transform_class=False):
+                 transform_class=False,
+                 center=True):
         self.zero_based = zero_based
         self.norm_type = norm_type
         self.transform_class = transform_class
+        self.center = center
 
     def __call__(self, data):
-
         dists = distribution.get_distributions(data)
         new_attrs = [self.normalize(dists[i], var) for
                      (i, var) in enumerate(data.domain.attributes)]
+
         new_class_vars = data.domain.class_vars
         if self.transform_class:
             attr_len = len(data.domain.attributes)
             new_class_vars = [self.normalize(dists[i + attr_len], var) for
                               (i, var) in enumerate(data.domain.class_vars)]
+
         domain = Domain(new_attrs, new_class_vars, data.domain.metas)
         return data.transform(domain)
 
@@ -41,7 +44,17 @@ def normalize_by_sd(self, dist, var):
         avg, sd = (dist.mean(), dist.standard_deviation()) if dist.size else (0, 1)
         if sd == 0:
             sd = 1
-        return ContinuousVariable(var.name, compute_value=Norm(var, avg, 1 / sd), sparse=var.sparse)
+
+        if self.center:
+            compute_val = Norm(var, avg, 1 / sd)
+        else:
+            compute_val = Norm(var, 0, 1 / sd)
+
+        return ContinuousVariable(
+            var.name,
+            compute_value=compute_val,
+            sparse=var.sparse,
+        )
 
     def normalize_by_span(self, dist, var):
         dma, dmi = dist.max(), dist.min()

Orange/preprocess/preprocess.py

@@ -269,6 +269,8 @@ class Normalize(Preprocess):
     Parameters
     ----------
     zero_based : bool (default=True)
+        Only used when `norm_type=NormalizeBySpan`.
+
         Determines the value used as the “low” value of the variable.
         It determines the interval for normalized continuous variables
         (either [-1, 1] or [0, 1]).
@@ -286,6 +288,11 @@ class Normalize(Preprocess):
     transform_class : bool (default=False)
         If True the class is normalized as well.
 
+    center : bool (default=True)
+        Only used when `norm_type=NormalizeBySD`.
+
+        Whether or not to center the data so it has mean zero.
+
     Examples
     --------
     >>> from Orange.data import Table
@@ -301,10 +308,12 @@ class Normalize(Preprocess):
     def __init__(self,
                  zero_based=True,
                  norm_type=NormalizeBySD,
-                 transform_class=False):
+                 transform_class=False,
+                 center=True):
         self.zero_based = zero_based
         self.norm_type = norm_type
         self.transform_class = transform_class
+        self.center = center
 
     def __call__(self, data):
         """
@@ -334,7 +343,9 @@ def __call__(self, data):
         normalizer = normalize.Normalizer(
             zero_based=self.zero_based,
             norm_type=self.norm_type,
-            transform_class=self.transform_class)
+            transform_class=self.transform_class,
+            center=self.center,
+        )
         return normalizer(data)
 
 

Orange/statistics/util.py

@@ -13,6 +13,23 @@
 from sklearn.utils.sparsefuncs import mean_variance_axis
 
 
+def sparse_array_equal(x1, x2):
+    """Check if two sparse arrays are equal."""
+    if not sp.issparse(x1):
+        raise TypeError("`x1` must be sparse.")
+    if not sp.issparse(x2):
+        raise TypeError("`x2` must be sparse.")
+
+    return x1.shape == x2.shape and (x1 != x2).nnz == 0
+
+
+def array_equal(x1, x2):
+    """Equivalent of np.array_equal that properly handles sparse matrices."""
+    if sp.issparse(x1) and sp.issparse(x2):
+        return sparse_array_equal(x1, x2)
+    return np.array_equal(x1, x2)
+
+
 def _count_nans_per_row_sparse(X, weights, dtype=None):
     """ Count the number of nans (undefined) values per row. """
     if weights is not None:

Orange/tests/test_statistics.py

@@ -9,7 +9,7 @@
 
 from Orange.statistics.util import bincount, countnans, contingency, digitize, \
     mean, nanmax, nanmean, nanmedian, nanmin, nansum, nanunique, stats, std, \
-    unique, var, nanstd, nanvar, nanmode
+    unique, var, nanstd, nanvar, nanmode, array_equal
 from sklearn.utils import check_random_state
 
 
@@ -590,6 +590,25 @@ def test_nanunique_ignores_nans_in_counts(self, array):
         np.testing.assert_equal(nanunique(x, return_counts=True)[1], expected)
 
 
+class TestArrayEqual(unittest.TestCase):
+    @dense_sparse
+    def test_same_matrices(self, array):
+        x = array([0, 1, 0, 0, 2])
+        self.assertTrue(array_equal(x, x))
+
+    @dense_sparse
+    def test_with_different_shapes(self, array):
+        x = array(np.eye(4))
+        y = array(np.eye(5))
+        self.assertFalse(array_equal(x, y))
+
+    @dense_sparse
+    def test_with_different_values(self, array):
+        x = array([0, 1, 0, 0, 2])
+        y = array([0, 3, 0, 0, 2])
+        self.assertFalse(array_equal(x, y))
+
+
 class TestNanModeAppVeyor(unittest.TestCase):
     def test_appveyour_still_not_onscipy_1_2_0(self):
         import scipy

Orange/widgets/data/tests/test_owfeaturestatistics.py

@@ -1,9 +1,9 @@
 import datetime
 import warnings
 from collections import namedtuple
-from functools import wraps, partial
+from functools import partial
 from itertools import chain
-from typing import Callable, List
+from typing import List
 
 import numpy as np
 from AnyQt.QtCore import QItemSelection, QItemSelectionRange, \
@@ -12,7 +12,7 @@
 from Orange.data import Table, Domain, StringVariable, ContinuousVariable, \
     DiscreteVariable, TimeVariable
 from Orange.widgets.tests.base import WidgetTest, datasets
-from Orange.widgets.tests.utils import simulate
+from Orange.widgets.tests.utils import simulate, table_dense_sparse
 from Orange.widgets.data.owfeaturestatistics import \
     OWFeatureStatistics
 
@@ -175,18 +175,6 @@ def make_table(attributes, target=None, metas=None):
     )
 
 
-def table_dense_sparse(test_case):
-    # type: (Callable) -> Callable
-    """Run a single test case on both dense and sparse Orange tables."""
-
-    @wraps(test_case)
-    def _wrapper(self):
-        test_case(self, lambda table: table.to_dense())
-        test_case(self, lambda table: table.to_sparse())
-
-    return _wrapper
-
-
 class TestVariousDataSets(WidgetTest):
     def setUp(self):
         self.widget = self.create_widget(

Orange/widgets/tests/utils.py

@@ -1,4 +1,6 @@
 import sys
+from functools import wraps
+
 import warnings
 import contextlib
 
@@ -317,3 +319,24 @@ def mouseMove(widget, pos=QPoint(), delay=-1):  # pragma: no-cover
         QTest.qWait(delay)
 
     QApplication.sendEvent(widget, me)
+
+
+def table_dense_sparse(test_case):
+    # type: (Callable) -> Callable
+    """Run a single test case on both dense and sparse Orange tables.
+
+    Examples
+    --------
+    >>> @table_dense_sparse
+    ... def test_something(self, prepare_table):
+    ...     data: Table  # The table you want to test on
+    ...     data = prepare_table(data)  # This converts the table to dense/sparse
+
+    """
+
+    @wraps(test_case)
+    def _wrapper(self):
+        test_case(self, lambda table: table.to_dense())
+        test_case(self, lambda table: table.to_sparse())
+
+    return _wrapper