Skip to content

TST add test_binomial_vs_alternative_formulation #26298

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 6 commits into from
May 16, 2023
Merged

TST add test_binomial_vs_alternative_formulation #26298

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
aae1a04
TST add test_binomial_vs_alternative_formulation
lorentzenchr Apr 24, 2023
0cac2b0
run CI again
lorentzenchr May 1, 2023
eef3ab4
CLN fix according to review
lorentzenchr May 2, 2023
4d3b487
CLN review comment
lorentzenchr May 3, 2023
b7894ca
CLN only one loop, keep generator
lorentzenchr May 4, 2023
99d3171
Updated test w/ parameterization
Micky774 May 10, 2023
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
36 changes: 36 additions & 0 deletions sklearn/_loss/tests/test_loss.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -981,6 +981,42 @@ def test_binomial_and_multinomial_loss(global_random_seed):
)


@pytest.mark.parametrize("y_true", (np.array([0.0, 0, 0]), np.array([1.0, 1, 1])))
@pytest.mark.parametrize("y_pred", (np.array([-5.0, -5, -5]), np.array([3.0, 3, 3])))
def test_binomial_vs_alternative_formulation(y_true, y_pred, global_dtype):
"""Test that both formulations of the binomial deviance agree.

Often, the binomial deviance or log loss is written in terms of a variable
z in {-1, +1}, but we use y in {0, 1}, hence z = 2 * y - 1.
ESL II Eq. (10.18):

-loglike(z, f) = log(1 + exp(-2 * z * f))

Note:
- ESL 2*f = raw_prediction, hence the factor 2 of ESL disappears.
- Deviance = -2*loglike + .., but HalfBinomialLoss is half of the
deviance, hence the factor of 2 cancels in the comparison.
"""

def alt_loss(y, raw_pred):
z = 2 * y - 1
return np.mean(np.log(1 + np.exp(-z * raw_pred)))

def alt_gradient(y, raw_pred):
# alternative gradient formula according to ESL
z = 2 * y - 1
return -z / (1 + np.exp(z * raw_pred))

bin_loss = HalfBinomialLoss()

y_true = y_true.astype(global_dtype)
y_pred = y_pred.astype(global_dtype)
datum = (y_true, y_pred)

assert bin_loss(*datum) == approx(alt_loss(*datum))
assert_allclose(bin_loss.gradient(*datum), alt_gradient(*datum))


@pytest.mark.parametrize("loss", LOSS_INSTANCES, ids=loss_instance_name)
def test_predict_proba(loss, global_random_seed):
"""Test that predict_proba and gradient_proba work as expected."""
Expand Down