Skip to content

TST ensure that sklearn/_loss/tests/test_loss.py is seed insensitive #22847

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 2 commits into from
Mar 15, 2022
Merged

TST ensure that sklearn/_loss/tests/test_loss.py is seed insensitive #22847

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
050c55a
TST ensure that sklearn/_loss/tests/test_loss.py is seed insensitive
ogrisel Mar 15, 2022
6c1c09f
Update sklearn/_loss/tests/test_loss.py
ogrisel Mar 15, 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
28 changes: 14 additions & 14 deletions sklearn/_loss/tests/test_loss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -412,7 +412,7 @@ def test_loss_same_as_C_functions(loss, sample_weight):

@pytest.mark.parametrize("loss", LOSS_INSTANCES, ids=loss_instance_name)
@pytest.mark.parametrize("sample_weight", [None, "range"])
def test_loss_gradients_are_the_same(loss, sample_weight):
def test_loss_gradients_are_the_same(loss, sample_weight, global_random_seed):
"""Test that loss and gradient are the same across different functions.

Also test that output arguments contain correct results.
Expand All @@ -422,7 +422,7 @@ def test_loss_gradients_are_the_same(loss, sample_weight):
n_samples=20,
y_bound=(-100, 100),
raw_bound=(-10, 10),
seed=42,
seed=global_random_seed,
)
if sample_weight == "range":
sample_weight = np.linspace(1, y_true.shape[0], num=y_true.shape[0])
Expand Down Expand Up @@ -493,7 +493,7 @@ def test_loss_gradients_are_the_same(loss, sample_weight):

@pytest.mark.parametrize("loss", LOSS_INSTANCES, ids=loss_instance_name)
@pytest.mark.parametrize("sample_weight", ["ones", "random"])
def test_sample_weight_multiplies(loss, sample_weight):
def test_sample_weight_multiplies(loss, sample_weight, global_random_seed):
"""Test sample weights in loss, gradients and hessians.

Make sure that passing sample weights to loss, gradient and hessian
Expand All @@ -505,13 +505,13 @@ def test_sample_weight_multiplies(loss, sample_weight):
n_samples=n_samples,
y_bound=(-100, 100),
raw_bound=(-5, 5),
seed=42,
seed=global_random_seed,
)

if sample_weight == "ones":
sample_weight = np.ones(shape=n_samples, dtype=np.float64)
else:
rng = np.random.RandomState(42)
rng = np.random.RandomState(global_random_seed)
sample_weight = rng.normal(size=n_samples).astype(np.float64)

assert_allclose(
Expand Down Expand Up @@ -635,7 +635,7 @@ def test_loss_of_perfect_prediction(loss, sample_weight):

@pytest.mark.parametrize("loss", LOSS_INSTANCES, ids=loss_instance_name)
@pytest.mark.parametrize("sample_weight", [None, "range"])
def test_gradients_hessians_numerically(loss, sample_weight):
def test_gradients_hessians_numerically(loss, sample_weight, global_random_seed):
"""Test gradients and hessians with numerical derivatives.

Gradient should equal the numerical derivatives of the loss function.
Expand All @@ -647,7 +647,7 @@ def test_gradients_hessians_numerically(loss, sample_weight):
n_samples=n_samples,
y_bound=(-100, 100),
raw_bound=(-5, 5),
seed=42,
seed=global_random_seed,
)

if sample_weight == "range":
Expand Down Expand Up @@ -799,7 +799,7 @@ def test_loss_intercept_only(loss, sample_weight):
if not loss.is_multiclass:
y_true = loss.link.inverse(np.linspace(-4, 4, num=n_samples))
else:
y_true = np.arange(n_samples).astype(float) % loss.n_classes
y_true = np.arange(n_samples).astype(np.float64) % loss.n_classes
y_true[::5] = 0 # exceedance of class 0

if sample_weight == "range":
Expand Down Expand Up @@ -867,13 +867,13 @@ def fun(x):
(HalfBinomialLoss(), np.mean, "binomial"),
],
)
def test_specific_fit_intercept_only(loss, func, random_dist):
def test_specific_fit_intercept_only(loss, func, random_dist, global_random_seed):
"""Test that fit_intercept_only returns the correct functional.

We test the functional for specific, meaningful distributions, e.g.
squared error estimates the expectation of a probability distribution.
"""
rng = np.random.RandomState(0)
rng = np.random.RandomState(global_random_seed)
if random_dist == "binomial":
y_train = rng.binomial(1, 0.5, size=100)
else:
Expand Down Expand Up @@ -921,9 +921,9 @@ def test_multinomial_loss_fit_intercept_only():
assert_all_finite(baseline_prediction)


def test_binomial_and_multinomial_loss():
def test_binomial_and_multinomial_loss(global_random_seed):
"""Test that multinomial loss with n_classes = 2 is the same as binomial loss."""
rng = np.random.RandomState(0)
rng = np.random.RandomState(global_random_seed)
n_samples = 20
binom = HalfBinomialLoss()
multinom = HalfMultinomialLoss(n_classes=2)
Expand All @@ -939,15 +939,15 @@ def test_binomial_and_multinomial_loss():


@pytest.mark.parametrize("loss", LOSS_INSTANCES, ids=loss_instance_name)
def test_predict_proba(loss):
def test_predict_proba(loss, global_random_seed):
"""Test that predict_proba and gradient_proba work as expected."""
n_samples = 20
y_true, raw_prediction = random_y_true_raw_prediction(
loss=loss,
n_samples=n_samples,
y_bound=(-100, 100),
raw_bound=(-5, 5),
seed=42,
seed=global_random_seed,
)

if hasattr(loss, "predict_proba"):
Expand Down