def test_kdebase(self):
data = np.array([[0, 2]])
tree = KDTree(data, leafsize=2)
query = np.array([[0, 2]])
tree_query = KDTree(query, leafsize=2)
result = tree.kdebase(tree.root, tree_query.root, tree_query, 1)
self.assertEqual([0.1592], round(result[0], 4))
def test_traverse_add(self):
data = np.array([[0, 2], [0, 3], [0, 4]])
tree = KDTree(data, leafsize=1)
lower = np.zeros(3)
upper = np.zeros(3)
l = 0.1
u = 0.2
tree.traverse_add(tree.root, lower, upper, l, u)
self.assertSequenceEqual(list([0.1, 0.1, 0.1]), list(lower))
self.assertSequenceEqual(list([0.2, 0.2, 0.2]), list(upper))
def test_LeafNode(self):
data = np.array([[0, 2]])
tree = KDTree(data, leafsize=2)
# Test if the node is recognized as leaf by the build function.
self.assertEqual(0, tree.root.idx)
self.assertSequenceEqual(list([0, 2]), list(tree.root.mins))
self.assertSequenceEqual(list([0, 2]), list(tree.root.maxes))
# Test if the information contained by the leaf node is correct.
self.assertEqual(1, len(tree.children(tree.root)))
self.assertEqual(0, tree.children(tree.root)[0].idx)
self.assertEqual(True, isinstance(tree.root, KDTree.LeafNode))
def test_build1(self):
data = np.array([[0, 2], [0, 3], [0, 4]])
tree = KDTree(data, leafsize=2)
# test if the node is built correctly.
self.assertSequenceEqual(list([0, 1, 2]), list(tree.root.idx))
self.assertSequenceEqual(list([0, 2]), list(tree.root.mins))
self.assertSequenceEqual(list([0, 4]), list(tree.root.maxes))
# Test if the information contained by the leaf node is correct.
self.assertSequenceEqual(list([0, 1]),
list(tree.children(tree.root)[0].idx))
self.assertSequenceEqual(list([2]),
list(tree.children(tree.root)[1].idx))
def test_if_change_low1(self):
query = np.array([[1, 2], [3, 4]])
tree_query = KDTree(query, leafsize=1)
data = np.array([[0, 2], [3, 4]])
tree = KDTree(data, leafsize=1)
self.assertTrue(not bound.if_change_low(
tree.root.mins[0], tree.root.maxes[0], tree_query.root.mins[0],
tree_query.root.maxes[0]))
self.assertTrue(not bound.if_change_low(
tree.root.mins[1], tree.root.maxes[1], tree_query.root.mins[1],
tree_query.root.maxes[1]))
def test_query_dual(self):
data = np.array([[1, 1]])
pts = np.array([[1, 1]])
tree = KDTree(data, leafsize=2)
treep = KDTree(pts, leafsize=2)
dens = KDTree.query_dual_tree(tree, treep, 1)
kfit = KernelDensity(bandwidth=1.0,
kernel="gaussian",
breadth_first=True,
leaf_size=2,
metric_params=None).fit(data)
result = kfit.score_samples(pts)
self.assertTrue((dens - np.exp(result)) < 0.001)
def test_query_dual_rand(self):
d = 5
data = np.random.rand(51, d)
pts = np.random.rand(3, d)
tree = KDTree(data, leafsize=4)
treep = KDTree(pts, leafsize=4)
dens = KDTree.query_dual_tree(tree, treep, 1)
kfit = KernelDensity(bandwidth=1.0,
kernel="gaussian",
breadth_first=True,
leaf_size=2,
metric_params=None).fit(data)
result = kfit.score_samples(pts)
self.assertTrue(np.sum(dens - np.exp(result)) < 0.001)
def test_find_bound(self):
query = np.array([[0, 2]])
tree_query = KDTree(query, leafsize=1)
data = np.array([[0, 2]])
tree = KDTree(data, leafsize=1)
# lower_distance = upper_distance = 0
l, u = bound.find_bound(tree.root, tree_query.root, 1)
self.assertEqual([0.1592], round(l, 4))
self.assertEqual([0.1592], round(u, 4))
query = np.array([[1, 2]])
tree_query = KDTree(query, leafsize=1)
data = np.array([[0, 2]])
tree = KDTree(data, leafsize=1)
l, u = bound.find_bound(tree.root, tree_query.root, 1)
self.assertEqual([0.1592], round(u, 4))
def test_children(self):
# LeafNode case
data = np.array([[0, 2]])
tree = KDTree(data, leafsize=2)
self.assertEqual(tree.children(tree.root)[0], tree.root)
# node case
data = np.array([[0, 2], [0, 3]])
tree = KDTree(data, leafsize=1)
self.assertEqual(tree.children(tree.root)[0].idx, [0])
self.assertEqual(tree.children(tree.root)[1].idx, [1])
def test_build2(self):
data = np.array([[1, 1, 2], [0, 1, 2], [3, 1, 2], [3, 2, 1], [3, 3, 4],
[3, 1, 3], [2, 3, 1], [3, 4, 1], [3, 1, 0], [3, 4,
5]])
tree = KDTree(data, leafsize=3)
# test if the node is built correctly on multi-dimensions.
self.assertSequenceEqual(list([0, 1, 0]), list(tree.root.mins))
self.assertSequenceEqual(list([3, 4, 5]), list(tree.root.maxes))
# split = 2.5; dimension = 3
# test if the left bound and right bound overlap
self.assertSequenceEqual(list([0, 1, 2, 3, 6, 7, 8]),
list(tree.root.leftchild.idx))
self.assertSequenceEqual(list([4, 5, 9]),
list(tree.root.rightchild.idx))
# show that after splitting, we still can use .idx to trace the value of the node
self.assertSequenceEqual(list([2, 3, 6, 7, 8]),
list(tree.root.leftchild.rightchild.idx))
self.assertSequenceEqual(list([0, 1]),
list(tree.root.leftchild.leftchild.idx))
# test if leave size == 3. root.greater shall be a LeafNode
self.assertTrue(isinstance(tree.root.rightchild, KDTree.LeafNode))
def test_traverse(self):
data = np.array([[0, 2], [0, 3]])
tree = KDTree(data, leafsize=1)
result = []
tree.traverse(tree.root, result) # result = array([0, 1])
self.assertSequenceEqual(list(result[0]), list([0, 1]))
def test_priority(self):
data = np.array([[0, 2], [0, 3]])
tree = KDTree(data, leafsize=1)
priority = tree.priority(tree.root, 0.1,
0.2) # 0.1 * 2 + 0.0001 * random(0, 1)
self.assertTrue((priority - 0.2) < 0.001)