def test_init_raw_predictions_shapes():
# Make sure get_init_raw_predictions returns float64 arrays with shape
# (n_samples, K) where K is 1 for binary classification and regression, and
# K = n_classes for multiclass classification
rng = np.random.RandomState(0)
n_samples = 100
X = rng.normal(size=(n_samples, 5))
y = rng.normal(size=n_samples)
for loss in (LeastSquaresError(n_classes=1),
LeastAbsoluteError(n_classes=1),
QuantileLossFunction(n_classes=1),
HuberLossFunction(n_classes=1)):
init_estimator = loss.init_estimator().fit(X, y)
raw_predictions = loss.get_init_raw_predictions(y, init_estimator)
assert raw_predictions.shape == (n_samples, 1)
assert raw_predictions.dtype == np.float64
y = rng.randint(0, 2, size=n_samples)
for loss in (BinomialDeviance(n_classes=2), ExponentialLoss(n_classes=2)):
init_estimator = loss.init_estimator().fit(X, y)
raw_predictions = loss.get_init_raw_predictions(y, init_estimator)
assert raw_predictions.shape == (n_samples, 1)
assert raw_predictions.dtype == np.float64
for n_classes in range(3, 5):
y = rng.randint(0, n_classes, size=n_samples)
loss = MultinomialDeviance(n_classes=n_classes)
init_estimator = loss.init_estimator().fit(X, y)
raw_predictions = loss.get_init_raw_predictions(y, init_estimator)
assert raw_predictions.shape == (n_samples, n_classes)
assert raw_predictions.dtype == np.float64
def test_lad_equals_quantiles(seed, alpha):
# Make sure quantile loss with alpha = .5 is equivalent to LAD
lad = LeastAbsoluteError()
ql = QuantileLossFunction(alpha=alpha)
n_samples = 50
rng = np.random.RandomState(seed)
raw_predictions = rng.normal(size=(n_samples))
y_true = rng.normal(size=(n_samples))
lad_loss = lad(y_true, raw_predictions)
ql_loss = ql(y_true, raw_predictions)
if alpha == 0.5:
assert lad_loss == approx(2 * ql_loss)
weights = np.linspace(0, 1, n_samples)**2
lad_weighted_loss = lad(y_true, raw_predictions, sample_weight=weights)
ql_weighted_loss = ql(y_true, raw_predictions, sample_weight=weights)
if alpha == 0.5:
assert lad_weighted_loss == approx(2 * ql_weighted_loss)
pbl_weighted_loss = mean_pinball_loss(y_true,
raw_predictions,
sample_weight=weights,
alpha=alpha)
assert pbl_weighted_loss == approx(ql_weighted_loss)
def test_lad_equals_quantile_50(seed):
# Make sure quantile loss with alpha = .5 is equivalent to LAD
lad = LeastAbsoluteError(n_classes=1)
ql = QuantileLossFunction(n_classes=1, alpha=0.5)
n_samples = 50
rng = np.random.RandomState(seed)
raw_predictions = rng.normal(size=(n_samples))
y_true = rng.normal(size=(n_samples))
lad_loss = lad(y_true, raw_predictions)
ql_loss = ql(y_true, raw_predictions)
assert_almost_equal(lad_loss, 2 * ql_loss)
weights = np.linspace(0, 1, n_samples)**2
lad_weighted_loss = lad(y_true, raw_predictions, sample_weight=weights)
ql_weighted_loss = ql(y_true, raw_predictions, sample_weight=weights)
assert_almost_equal(lad_weighted_loss, 2 * ql_weighted_loss)