def test_kmean_nonsense_centroid(self):
centroids = [[1.0, 1.0, 1.0], [2.0, 2.0, 2.0], [3.0, 3.0, 3.0], [99999.0, 99999.0, 99999.0]]
expected_centroids = [[1.0, 1.0, 1.0], [2.0, 2.0, 2.0], [3.0, 3.0, 3.0]]
clean_points = [p for p in int_points_3d if 100 not in p]
# Since we pass a 4th centroid guess that is completely ridiculous, we
# expect that the resulting centroids will only contain 3 members.
m_centroids, _ = kmeans.kmeans(clean_points, centroids)
self.assertSequenceEqual(expected_centroids, m_centroids)
def test_kmean_nonsense_centroid(self):
centroids = [[1., 1., 1.], [2., 2., 2.], [3., 3., 3.],
[99999., 99999., 99999.]]
expected_centroids = [[1., 1., 1.], [2., 2., 2.], [3., 3., 3.]]
clean_points = [p for p in int_points_3d if 100 not in p]
# Since we pass a 4th centroid guess that is completely ridiculous, we
# expect that the resulting centroids will only contain 3 members.
m_centroids, _ = kmeans.kmeans(clean_points, centroids)
self.assertSequenceEqual(expected_centroids, m_centroids)
def test_kmeans(self):
# These indices don't really matter since the points are random but
# I am fixing them here for repeatability of the test.
centroids = [random_points_3d[125],
random_points_3d[500],
random_points_3d[875]]
s_centroids = [[ 0.44876204, 0.3331773 , 0.233552 ],
[ 0.49838519, 0.29378851, 0.75018887],
[ 0.5225907 , 0.80407079, 0.53268326]]
m_centroids, m_distance = kmeans.kmeans(random_points_3d, centroids)
self.assertEqual(0.35944390987038655, m_distance)
# I'm getting everything to pass at 10 places except for this where
# there always seems to be at least one dimension of one centroid
# about [.001-.005] away from where we expect it to be. This is due to
# difference in floating point storage between numpy and python. See
# the following for more info:
# https://github.com/cbmi/avocado/issues/34
self.assertSequenceAlmostEqual(s_centroids, m_centroids, num_places=2)
self.assertRaises(ValueError, kmeans.kmeans, [], 3)
self.assertRaises(ValueError, kmeans.kmeans, random_points_3d, 0)
def test_kmeans(self):
# These indices don't really matter since the points are random but
# I am fixing them here for repeatability of the test.
centroids = [
random_points_3d[125], random_points_3d[500], random_points_3d[875]
]
s_centroids = [[0.44876204, 0.3331773, 0.233552],
[0.49838519, 0.29378851, 0.75018887],
[0.5225907, 0.80407079, 0.53268326]]
m_centroids, m_distance = kmeans.kmeans(random_points_3d, centroids)
self.assertEqual(0.35944390987038655, m_distance)
# I'm getting everything to pass at 10 places except for this where
# there always seems to be at least one dimension of one centroid
# about [.001-.005] away from where we expect it to be. This is due to
# difference in floating point storage between numpy and python. See
# the following for more info:
# https://github.com/cbmi/avocado/issues/34
self.assertSequenceAlmostEqual(s_centroids, m_centroids, num_places=2)
self.assertRaises(ValueError, kmeans.kmeans, [], 3)
self.assertRaises(ValueError, kmeans.kmeans, random_points_3d, 0)