Skip to content

TST use global_dtype in sklearn/feature_selection/tests/test_mutual_info.py #22677

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 13 commits into from
Mar 25, 2022
Merged

TST use global_dtype in sklearn/feature_selection/tests/test_mutual_info.py #22677

Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
61a1966
TST Adapt test_mutual_info.py to test implementations on 32bit datasets
jjerphan Mar 3, 2022
8712833
Merge branch 'main' into tst/test_mutual_info-32bit
jjerphan Mar 9, 2022
38b1c65
Apply comments from review
jjerphan Mar 9, 2022
57acb77
Merge branch 'main' into tst/test_mutual_info-32bit
jjerphan Mar 17, 2022
bba8a12
TST Use global_dtype
jjerphan Mar 17, 2022
33cc917
TST Review comments
jjerphan Mar 18, 2022
61022e1
Address comments
jjerphan Mar 23, 2022
b31cccb
TST Adapt test_mean_shift.py to test implementations on 32bit datasets
jjerphan Mar 3, 2022
ab34666
TST Use global_dtype
jjerphan Mar 17, 2022
70cde82
Address review comments
jjerphan Mar 17, 2022
063b935
Assert review comments
jjerphan Mar 23, 2022
8a95627
TST Undo changes made to other files
jjerphan Mar 23, 2022
b7b33e8
Merge branch 'main' into tst/test_mutual_info-32bit
jeremiedbb Mar 25, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
60 changes: 35 additions & 25 deletions sklearn/feature_selection/tests/test_mutual_info.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@
from sklearn.utils import check_random_state
from sklearn.utils._testing import (
assert_array_equal,
assert_almost_equal,
assert_allclose,
)
from sklearn.feature_selection._mutual_info import _compute_mi
Expand All @@ -22,7 +21,7 @@ def test_compute_mi_dd():
H_xy = -1 / 5 * np.log(1 / 5) - 2 / 5 * np.log(2 / 5) - 2 / 5 * np.log(2 / 5)
I_xy = H_x + H_y - H_xy

assert_almost_equal(_compute_mi(x, y, True, True), I_xy)
assert_allclose(_compute_mi(x, y, x_discrete=True, y_discrete=True), I_xy)


def test_compute_mi_cc(global_dtype):
Expand Down Expand Up @@ -54,11 +53,13 @@ def test_compute_mi_cc(global_dtype):
# Theory and computed values won't be very close
# We here check with a large relative tolerance
for n_neighbors in [3, 5, 7]:
I_computed = _compute_mi(x, y, False, False, n_neighbors)
I_computed = _compute_mi(
x, y, x_discrete=False, y_discrete=False, n_neighbors=n_neighbors
)
assert_allclose(I_computed, I_theory, rtol=1e-1)


def test_compute_mi_cd():
def test_compute_mi_cd(global_dtype):
# To test define a joint distribution as follows:
# p(x, y) = p(x) p(y | x)
# X ~ Bernoulli(p)
Expand All @@ -80,7 +81,7 @@ def test_compute_mi_cd():
for p in [0.3, 0.5, 0.7]:
x = rng.uniform(size=n_samples) > p

y = np.empty(n_samples)
y = np.empty(n_samples, global_dtype)
mask = x == 0
y[mask] = rng.uniform(-1, 1, size=np.sum(mask))
y[~mask] = rng.uniform(0, 2, size=np.sum(~mask))
Expand All @@ -91,32 +92,36 @@ def test_compute_mi_cd():

# Assert the same tolerance.
for n_neighbors in [3, 5, 7]:
I_computed = _compute_mi(x, y, True, False, n_neighbors)
assert_almost_equal(I_computed, I_theory, 1)
I_computed = _compute_mi(
x, y, x_discrete=True, y_discrete=False, n_neighbors=n_neighbors
)
assert_allclose(I_computed, I_theory, rtol=1e-1)


def test_compute_mi_cd_unique_label():
def test_compute_mi_cd_unique_label(global_dtype):
# Test that adding unique label doesn't change MI.
n_samples = 100
x = np.random.uniform(size=n_samples) > 0.5

y = np.empty(n_samples)
y = np.empty(n_samples, global_dtype)
mask = x == 0
y[mask] = np.random.uniform(-1, 1, size=np.sum(mask))
y[~mask] = np.random.uniform(0, 2, size=np.sum(~mask))

mi_1 = _compute_mi(x, y, True, False)
mi_1 = _compute_mi(x, y, x_discrete=True, y_discrete=False)

x = np.hstack((x, 2))
y = np.hstack((y, 10))
mi_2 = _compute_mi(x, y, True, False)
mi_2 = _compute_mi(x, y, x_discrete=True, y_discrete=False)

assert mi_1 == mi_2
assert_allclose(mi_1, mi_2)


# We are going test that feature ordering by MI matches our expectations.
def test_mutual_info_classif_discrete():
X = np.array([[0, 0, 0], [1, 1, 0], [2, 0, 1], [2, 0, 1], [2, 0, 1]])
def test_mutual_info_classif_discrete(global_dtype):
X = np.array(
[[0, 0, 0], [1, 1, 0], [2, 0, 1], [2, 0, 1], [2, 0, 1]], dtype=global_dtype
)
y = np.array([0, 1, 2, 2, 1])

# Here X[:, 0] is the most informative feature, and X[:, 1] is weakly
Expand All @@ -125,7 +130,7 @@ def test_mutual_info_classif_discrete():
assert_array_equal(np.argsort(-mi), np.array([0, 2, 1]))


def test_mutual_info_regression():
def test_mutual_info_regression(global_dtype):
# We generate sample from multivariate normal distribution, using
# transformation from initially uncorrelated variables. The zero
# variables after transformation is selected as the target vector,
Expand All @@ -136,19 +141,22 @@ def test_mutual_info_regression():
mean = np.zeros(4)

rng = check_random_state(0)
Z = rng.multivariate_normal(mean, cov, size=1000)
Z = rng.multivariate_normal(mean, cov, size=1000).astype(global_dtype, copy=False)
X = Z[:, 1:]
y = Z[:, 0]

mi = mutual_info_regression(X, y, random_state=0)
assert_array_equal(np.argsort(-mi), np.array([1, 2, 0]))
# XXX: should mutual_info_regression be fixed to avoid
# up-casting float32 inputs to float64?
assert mi.dtype == np.float64


def test_mutual_info_classif_mixed():
def test_mutual_info_classif_mixed(global_dtype):
# Here the target is discrete and there are two continuous and one
# discrete feature. The idea of this test is clear from the code.
rng = check_random_state(0)
X = rng.rand(1000, 3)
X = rng.rand(1000, 3).astype(global_dtype, copy=False)
X[:, 1] += X[:, 0]
y = ((0.5 * X[:, 0] + X[:, 2]) > 0.5).astype(int)
X[:, 2] = X[:, 2] > 0.5
Expand All @@ -168,9 +176,11 @@ def test_mutual_info_classif_mixed():
assert mi_nn[2] == mi[2]


def test_mutual_info_options():
X = np.array([[0, 0, 0], [1, 1, 0], [2, 0, 1], [2, 0, 1], [2, 0, 1]], dtype=float)
y = np.array([0, 1, 2, 2, 1], dtype=float)
def test_mutual_info_options(global_dtype):
X = np.array(
[[0, 0, 0], [1, 1, 0], [2, 0, 1], [2, 0, 1], [2, 0, 1]], dtype=global_dtype
)
y = np.array([0, 1, 2, 2, 1], dtype=global_dtype)
X_csr = csr_matrix(X)

for mutual_info in (mutual_info_regression, mutual_info_classif):
Expand All @@ -192,8 +202,8 @@ def test_mutual_info_options():
mi_5 = mutual_info(X, y, discrete_features=[True, False, True], random_state=0)
mi_6 = mutual_info(X, y, discrete_features=[0, 2], random_state=0)

assert_array_equal(mi_1, mi_2)
assert_array_equal(mi_3, mi_4)
assert_array_equal(mi_5, mi_6)
assert_allclose(mi_1, mi_2)
assert_allclose(mi_3, mi_4)
assert_allclose(mi_5, mi_6)

assert not np.allclose(mi_1, mi_3)
assert not np.allclose(mi_1, mi_3)