def test__predict_by_tree():
np.random.seed(42)
X = np.reshape(np.arange(0, 40), (10, 4))
Y = np.random.randn(10, 1)
depth_max = 4
size_min_leaf = 2
num_features = 2
split_random_location = True
node = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node, depth_max, size_min_leaf, num_features,
split_random_location, 1)
with pytest.raises(AssertionError) as error:
package_target._predict_by_tree(np.array([4.0, 2.0, 3.0, 1.0]), 'abc')
with pytest.raises(AssertionError) as error:
package_target._predict_by_tree(X, node)
with pytest.raises(AssertionError) as error:
package_target._predict_by_tree('abc', node)
mean, std = package_target._predict_by_tree(np.array([4.0, 2.0, 3.0, 1.0]),
node)
print(mean)
print(std)
assert mean == 0.179224925920024
assert std == 0.31748922709120864
def get_random_forest(
X: np.ndarray,
Y: np.ndarray,
num_trees: int,
depth_max: int,
size_min_leaf: int,
num_features: int,
) -> list:
"""
It returns a random forest.
:param X: inputs. Shape: (N, d).
:type X: np.ndarray
:param Y: outputs. Shape: (N, 1).
:type Y: str.
:param num_trees: the number of trees.
:type num_trees: int.
:param depth_max: maximum depth of tree.
:type depth_max: int.
:param size_min_leaf: minimum size of leaf.
:type size_min_leaf: int.
:param num_features: the number of split features.
:type num_features: int.
:returns: list of trees
:rtype: list
:raises: AssertionError
"""
assert isinstance(X, np.ndarray)
assert isinstance(Y, np.ndarray)
assert isinstance(num_trees, int)
assert isinstance(depth_max, int)
assert isinstance(size_min_leaf, int)
assert isinstance(num_features, int)
assert len(X.shape) == 2
assert len(Y.shape) == 2
assert X.shape[0] == Y.shape[0]
assert Y.shape[1] == 1
ratio_sample = 1.0
replace_samples = True
split_random_location = False
list_trees = []
for _ in range(0, num_trees):
X_, Y_ = trees_common.subsample(X, Y, ratio_sample, replace_samples)
root = trees_common._split(X_, Y_, num_features, split_random_location)
trees_common.split(root, depth_max, size_min_leaf, num_features,
split_random_location, 1)
list_trees.append(root)
return list_trees
def test__predict_by_trees():
np.random.seed(42)
X = np.reshape(np.arange(0, 40), (10, 4))
Y = np.random.randn(10, 1)
depth_max = 4
size_min_leaf = 2
num_features = 2
split_random_location = True
node_1 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_1, depth_max, size_min_leaf, num_features,
split_random_location, 1)
node_2 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_2, depth_max, size_min_leaf, num_features,
split_random_location, 1)
node_3 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_3, depth_max, size_min_leaf, num_features,
split_random_location, 1)
list_trees = [node_1, node_2, node_3]
with pytest.raises(AssertionError) as error:
package_target._predict_by_trees(np.array([4.0, 2.0, 3.0, 1.0]),
node_1)
with pytest.raises(AssertionError) as error:
package_target._predict_by_trees(np.array([4.0, 2.0, 3.0, 1.0]), 'abc')
with pytest.raises(AssertionError) as error:
package_target._predict_by_trees(X, list_trees)
with pytest.raises(AssertionError) as error:
package_target._predict_by_trees('abc', list_trees)
mean, std = package_target._predict_by_trees(
np.array([4.0, 2.0, 3.0, 1.0]), list_trees)
print(mean)
print(std)
assert mean == 0.12544669602080652
assert std == 0.3333040901154691
def test_split():
np.random.seed(42)
X = np.reshape(np.arange(0, 40), (10, 4))
Y = np.random.randn(10, 1)
depth_max = 4
size_min_leaf = 2
num_features = 2
split_random_location = False
node = package_target._split(X, Y, num_features, split_random_location)
with pytest.raises(AssertionError) as error:
package_target.split(node, depth_max, size_min_leaf, num_features,
split_random_location, 'abc')
with pytest.raises(AssertionError) as error:
package_target.split(node, depth_max, size_min_leaf, num_features,
split_random_location, 1.0)
with pytest.raises(AssertionError) as error:
package_target.split(node, depth_max, size_min_leaf, num_features,
'abc', 1)
with pytest.raises(AssertionError) as error:
package_target.split(node, depth_max, size_min_leaf, 'abc',
split_random_location, 1)
with pytest.raises(AssertionError) as error:
package_target.split(node, depth_max, 'abc', num_features,
split_random_location, 1)
with pytest.raises(AssertionError) as error:
package_target.split(node, 'abc', size_min_leaf, num_features,
split_random_location, 1)
with pytest.raises(AssertionError) as error:
package_target.split(X, depth_max, size_min_leaf, num_features,
split_random_location, 1)
with pytest.raises(AssertionError) as error:
package_target.split('abc', depth_max, size_min_leaf, num_features,
split_random_location, 1)
package_target.split(node, depth_max, size_min_leaf, num_features,
split_random_location, 1)
assert isinstance(node, dict)
def get_generic_trees(
X: np.ndarray,
Y: np.ndarray,
num_trees: int,
depth_max: int,
size_min_leaf: int,
ratio_sampling: float,
replace_samples: bool,
num_features: int,
split_random_location: bool,
) -> list:
"""
It returns a list of generic trees.
:param X: inputs. Shape: (N, d).
:type X: np.ndarray
:param Y: outputs. Shape: (N, 1).
:type Y: str.
:param num_trees: the number of trees.
:type num_trees: int.
:param depth_max: maximum depth of tree.
:type depth_max: int.
:param size_min_leaf: minimum size of leaf.
:type size_min_leaf: int.
:param ratio_sampling: ratio of dataset subsampling.
:type ratio_sampling: float
:param replace_samples: flag for replacement.
:type replace_samples: bool.
:param num_features: the number of split features.
:type num_features: int.
:param split_random_location: flag for random split location.
:type split_random_location: bool.
:returns: list of trees
:rtype: list
:raises: AssertionError
"""
assert isinstance(X, np.ndarray)
assert isinstance(Y, np.ndarray)
assert isinstance(num_trees, int)
assert isinstance(depth_max, int)
assert isinstance(size_min_leaf, int)
assert isinstance(ratio_sampling, float)
assert isinstance(replace_samples, bool)
assert isinstance(num_features, int)
assert isinstance(split_random_location, bool)
assert len(X.shape) == 2
assert len(Y.shape) == 2
assert X.shape[0] == Y.shape[0]
assert Y.shape[1] == 1
if replace_samples:
assert ratio_sampling > 0.0
else:
assert 0.0 < ratio_sampling <= 1.0
list_trees = []
for _ in range(0, num_trees):
X_, Y_ = trees_common.subsample(X, Y, ratio_sampling, replace_samples)
root = trees_common._split(X_, Y_, num_features, split_random_location)
trees_common.split(root, depth_max, size_min_leaf, num_features,
split_random_location, 1)
list_trees.append(root)
return list_trees
def test_predict_by_trees():
np.random.seed(42)
X = np.reshape(np.arange(0, 40), (10, 4))
Y = np.random.randn(10, 1)
depth_max = 4
size_min_leaf = 2
num_features = 2
split_random_location = True
node_1 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_1, depth_max, size_min_leaf, num_features,
split_random_location, 1)
node_2 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_2, depth_max, size_min_leaf, num_features,
split_random_location, 1)
node_3 = package_target._split(X, Y, num_features, split_random_location)
package_target.split(node_3, depth_max, size_min_leaf, num_features,
split_random_location, 1)
list_trees = [node_1, node_2, node_3]
with pytest.raises(AssertionError) as error:
package_target.predict_by_trees(np.array([4.0, 2.0, 3.0, 1.0]), node_1)
with pytest.raises(AssertionError) as error:
package_target.predict_by_trees(np.array([4.0, 2.0, 3.0, 1.0]), 'abc')
with pytest.raises(AssertionError) as error:
package_target.predict_by_trees(np.array([4.0, 2.0, 3.0, 1.0]),
list_trees)
with pytest.raises(AssertionError) as error:
package_target.predict_by_trees('abc', list_trees)
means, stds = package_target.predict_by_trees(X, list_trees)
print(means)
print(stds)
means_truth = np.array([
[0.33710618],
[0.1254467],
[0.68947573],
[0.9866563],
[0.16257799],
[0.16258346],
[0.85148454],
[0.55084716],
[0.30721653],
[0.30721653],
])
stds_truth = np.array([
[0.29842457],
[0.33330409],
[0.44398864],
[0.72536523],
[0.74232577],
[0.74232284],
[0.83388663],
[0.5615399],
[0.64331582],
[0.64331582],
])
assert isinstance(means, np.ndarray)
assert isinstance(stds, np.ndarray)
assert len(means.shape) == 2
assert len(stds.shape) == 2
assert means.shape[0] == stds.shape[0] == X.shape[0]
assert means.shape[1] == stds.shape[1] == 1
assert np.all(np.abs(means - means_truth) < TEST_EPSILON)
assert np.all(np.abs(stds - stds_truth) < TEST_EPSILON)
X = np.random.randn(1000, 4)
means, stds = package_target.predict_by_trees(X, list_trees)
assert isinstance(means, np.ndarray)
assert isinstance(stds, np.ndarray)
assert len(means.shape) == 2
assert len(stds.shape) == 2
assert means.shape[0] == stds.shape[0] == X.shape[0]
assert means.shape[1] == stds.shape[1] == 1
def test__split():
np.random.seed(42)
X = np.reshape(np.arange(0, 40), (10, 4))
Y = np.random.randn(10, 1)
num_features = 2
split_random_location = False
with pytest.raises(AssertionError) as error:
package_target._split(X, Y, num_features, 'abc')
with pytest.raises(AssertionError) as error:
package_target._split(X, Y, 'abc', split_random_location)
with pytest.raises(AssertionError) as error:
package_target._split(X, Y, 2.0, split_random_location)
with pytest.raises(AssertionError) as error:
package_target._split(X, 'abc', num_features, split_random_location)
with pytest.raises(AssertionError) as error:
package_target._split('abc', Y, num_features, split_random_location)
dict_split = package_target._split(X, Y, num_features,
split_random_location)
print(dict_split)
print(dict_split['index'])
print(dict_split['value'])
print(dict_split['left_right'])
assert isinstance(dict_split, dict)
assert dict_split['index'] == 1
assert dict_split['value'] == 35.0
assert np.all(dict_split['left_right'][0][0][0] == np.array([0, 1, 2, 3]))
assert np.abs(dict_split['left_right'][0][0][1] -
np.array([0.49671415])) < TEST_EPSILON
assert np.all(dict_split['left_right'][0][1][0] == np.array([4, 5, 6, 7]))
assert np.abs(dict_split['left_right'][0][1][1] -
np.array([-0.1382643])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][2][0] == np.array([8, 9, 10, 11]))
assert np.abs(dict_split['left_right'][0][2][1] -
np.array([0.64768854])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][3][0] == np.array([12, 13, 14, 15]))
assert np.abs(dict_split['left_right'][0][3][1] -
np.array([1.52302986])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][4][0] == np.array([16, 17, 18, 19]))
assert np.abs(dict_split['left_right'][0][4][1] -
np.array([-0.23415337])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][5][0] == np.array([20, 21, 22, 23]))
assert np.abs(dict_split['left_right'][0][5][1] -
np.array([-0.23413696])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][6][0] == np.array([24, 25, 26, 27]))
assert np.abs(dict_split['left_right'][0][6][1] -
np.array([1.57921282])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][7][0] == np.array([28, 29, 30, 31]))
assert np.abs(dict_split['left_right'][0][7][1] -
np.array([0.76743473])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][0][8][0] == np.array([32, 33, 34, 35]))
assert np.abs(dict_split['left_right'][0][8][1] -
np.array([-0.46947439])) < TEST_EPSILON
assert np.all(
dict_split['left_right'][1][0][0] == np.array([36, 37, 38, 39]))
assert np.abs(dict_split['left_right'][1][0][1] -
np.array([0.54256004])) < TEST_EPSILON
dict_split = package_target._split(X, Y, num_features, True)
print(dict_split)
print(dict_split['index'])
print(dict_split['value'])
print(dict_split['left_right'])
assert isinstance(dict_split, dict)
assert dict_split['index'] == 0
assert dict_split['value'] == 0.2543869879098266
X = np.ones(X.shape)
dict_split = package_target._split(X, Y, num_features, True)
print(dict_split)
print(dict_split['index'])
print(dict_split['value'])
print(dict_split['left_right'])
assert isinstance(dict_split, dict)
assert dict_split['index'] == 1
assert dict_split['value'] == 1.0