def test_try_include(self):
""" Tests if try_include detects a used point and empty cluster_set.
Tests that the point gets included when threshold = INFINITY.
Tests that the point is disregarded when threshold = -1
-------
"""
used = numpy.ones(DIMENSIONS)
used[0] = numpy.nan
self.assertTrue(setlib.try_include(used, arbitrary_set, model),
"Used point not detected")
self.assertFalse(setlib.try_include(point, [], model),
"Empty set not detected")
model.threshold = -1
self.assertFalse(setlib.try_include(point, arbitrary_set, model),
"Incorrectly included")
model.threshold = INFINITY
self.assertTrue(setlib.try_include(point, arbitrary_set, model),
"Incorrectly excluded")
self.assertGreater(arbitrary_set[0].size, 1, "Set not updated")
arbitrary_set[0] = clustlib.Cluster(DIMENSIONS)
clustlib.update_cluster(point, arbitrary_set[0])
model.threshold = model.eucl_threshold
def test_variance(self):
""" Tests that the has_variance flag of a cluster updates accordingly.
-------
"""
cluster = clustlib.Cluster(2)
clustlib.update_cluster(numpy.ones(2), cluster)
self.assertFalse(cluster.has_variance,
"Cluster has variance when it should not")
clustlib.update_cluster(numpy.ones(2) * 2, cluster)
self.assertTrue(cluster.has_variance, "Cluster variance not updated")
def test_has_variance(self):
""" Tests that has_variance only evaluates to true when it has a non zero std_dev in
each dimension
-------
"""
no_variance = clustlib.has_variance(ones)
cluster = clustlib.Cluster(DIMENSIONS)
other = clustlib.Cluster(DIMENSIONS)
clustlib.update_cluster(point, cluster)
clustlib.update_cluster(point, other)
point_a = numpy.ones(2)
point_b = numpy.ones(2)
point_a[1] = 2
point_b[0] = 2
clustlib.update_cluster(point_a, cluster)
clustlib.update_cluster(point_b, other)
merged = clustlib.merge_clusters(other, cluster)
self.assertFalse(no_variance, "No dimension has variance")
self.assertFalse(clustlib.has_variance(cluster),
"Second dimension has variance")
self.assertFalse(clustlib.has_variance(other),
"First dimension has variance")
self.assertTrue(clustlib.has_variance(merged),
"Both dimensions have variance")
def test_cluster_point(self):
""" Tests that the closest cluster is updated and that a used point is detected.
-------
"""
another = clustlib.Cluster(2)
clustlib.update_cluster(point, another)
model.discard.append(another)
model.discard.append(twos)
clustlib.cluster_point(point, model)
self.assertEqual(model.discard[0].size, 2)
nan = numpy.zeros(2)
nan[0] = numpy.nan
clustlib.cluster_point(nan, model)
model.discard = []
def test_update_cluster(self):
""" Tests that a cluster get its attributes updated accordingly.
-------
"""
merged = clustlib.merge_clusters(ones, ones)
size_b4 = merged.size
sums_b4 = merged.sums[0]
sums_sq_b4 = merged.sums_sq[0]
self.assertFalse(clustlib.has_variance(merged),
"Cluster has no variance")
for i in range(5):
i += 1
clustlib.update_cluster(other_point, merged)
self.assertEqual(size_b4 + i, merged.size, "Incorrect size")
self.assertEqual(sums_b4 + i * 2, merged.sums[0], "Incorrect sum")
self.assertEqual(sums_sq_b4 + (i * 4), merged.sums_sq[0],
"incorrect sums_sq")
self.assertTrue(merged.has_variance, "Cluster has variance")
def test_std_dev(self):
""" Tests that two computed population standard deviations compute to the same value
as when computing with numpy
-------
"""
yet_another = numpy.ones(2)
yet_another = yet_another * 4
fours = clustlib.Cluster(2)
clustlib.update_cluster(yet_another, fours)
merged = clustlib.merge_clusters(fours, ones)
std_dev = clustlib.std_dev(merged)
self.assertEqual(std_dev[0], numpy.std([1, 4]), "Incorrect std_dev")
clustlib.update_cluster(point * 5, merged)
std_dev = clustlib.std_dev(merged)
res = numpy.std([1, 4, 5])
std_dev[1] = round(std_dev[1], 5)
res = round(res, 5)
self.assertEqual(std_dev[1], res, "Incorrect std_dev")
DIMENSIONS = 2
NOF_POINTS = 5
NOF_CLUSTERS = 5
model = bfr.Model(mahalanobis_factor=3.0,
euclidean_threshold=5.0,
merge_threshold=10.0,
dimensions=DIMENSIONS,
init_rounds=10,
nof_clusters=NOF_CLUSTERS)
model.initialized = True
INFINITY = 13371337.0
point = numpy.ones(2)
other_point = point * 2
ones = clustlib.Cluster(2)
twos = clustlib.Cluster(2)
clustlib.update_cluster(point, ones)
clustlib.update_cluster(other_point, twos)
arbitrary_set = [ones, ones]
class PtlibTests(unittest.TestCase):
"""Test cases for the module bfr.ptlib"""
class PtlibTests(unittest.TestCase):
def test_squared_diff(self):
""" Tests that the sum of dimensions of ({1, 1} - {2, 2}) ^ 2 == 2
-------