def test_multiple_outputs_balanced_weights_cv(self):
"""Test MultiTaskWeightedLassoCV with weights."""
# Define weights
sample_weight = np.concatenate(
(np.ones(TestLassoExtensions.n_samples // 2),
np.ones(TestLassoExtensions.n_samples // 2) * 2))
# Define extended datasets
X_expanded = np.concatenate(
(TestLassoExtensions.X,
TestLassoExtensions.X[TestLassoExtensions.n_samples // 2:]))
y_expanded = np.concatenate(
(TestLassoExtensions.y_2D,
TestLassoExtensions.y_2D[TestLassoExtensions.n_samples // 2:]))
# Define splitters
# WeightedKFold splitter
cv_splitter = WeightedKFold(n_splits=3)
wlasso_cv = list(
cv_splitter.split(TestLassoExtensions.X,
TestLassoExtensions.y_2D,
sample_weight=sample_weight))
# Map weighted splitter to an extended splitter
index_mapper = {}
for i in range(TestLassoExtensions.n_samples):
if i < TestLassoExtensions.n_samples // 2:
index_mapper[i] = [i]
else:
index_mapper[i] = [i, i + TestLassoExtensions.n_samples // 2]
lasso_cv = self._map_splitter(
wlasso_cv,
TestLassoExtensions.n_samples + TestLassoExtensions.n_samples // 2,
index_mapper)
# Define alphas to test
alphas = np.logspace(-4, -1, num=10)
# Compare with LassoCV
# --> No intercept
params = {'fit_intercept': False}
self._compare_with_lasso_cv(X_expanded,
y_expanded,
TestLassoExtensions.X,
TestLassoExtensions.y_2D,
sample_weight=sample_weight,
alphas=alphas,
lasso_cv=lasso_cv,
wlasso_cv=wlasso_cv,
params=params)
# --> With intercept
params = {'fit_intercept': True}
self._compare_with_lasso_cv(X_expanded,
y_expanded,
TestLassoExtensions.X,
TestLassoExtensions.y_2D,
sample_weight=sample_weight,
alphas=alphas,
lasso_cv=lasso_cv,
wlasso_cv=wlasso_cv,
params=params)
def test_weighted_KFold(self):
"""Test WeightedKFold used in WeightedLassoCV."""
# Choose a smaller n to speed-up process
n = 100
sample_weight = np.random.choice(10, size=n)
n_splits = 3
wkf = WeightedKFold(n_splits=n_splits)
total_weight = np.sum(sample_weight)
for _, test_index in wkf.split(TestLassoExtensions.X[:n],
TestLassoExtensions.y_simple[:n], sample_weight=sample_weight):
# Compare fold weights
self.assertAlmostEqual(
np.sum(sample_weight[test_index]) / total_weight, 1 / n_splits,
delta=5e-2)
def test_multiple_outputs_no_weights_cv(self):
"""Test MultiTaskWeightedLassoCV with no weights."""
# Define alphas to test
alphas = np.logspace(-4, -1, num=10)
# Define splitter
cv = WeightedKFold(n_splits=3)
# Compare with MultiTaskLassoCV
# --> No intercept
params = {'fit_intercept': False}
self._compare_with_lasso_cv(TestLassoExtensions.X,
TestLassoExtensions.y_2D,
TestLassoExtensions.X,
TestLassoExtensions.y_2D,
sample_weight=None,
alphas=alphas,
lasso_cv=cv,
wlasso_cv=cv,
params=params)
# --> With intercept
params = {'fit_intercept': True}
self._compare_with_lasso_cv(TestLassoExtensions.X,
TestLassoExtensions.y_2D,
TestLassoExtensions.X,
TestLassoExtensions.y_2D,
sample_weight=None,
alphas=alphas,
lasso_cv=cv,
wlasso_cv=cv,
params=params)
def _weighted_check_cv(cv='warn', y=None, classifier=False):
if cv is None or cv == 'warn':
warnings.warn(CV_WARNING, FutureWarning)
cv = 3
if isinstance(cv, numbers.Integral):
if (classifier and (y is not None)
and (type_of_target(y) in ('binary', 'multiclass'))):
return WeightedStratifiedKFold(cv)
else:
return WeightedKFold(cv)
if not hasattr(cv, 'split') or isinstance(cv, str):
if not isinstance(cv, Iterable) or isinstance(cv, str):
raise ValueError("Expected cv as an integer, cross-validation "
"object (from sklearn.model_selection) "
"or an iterable. Got %s." % cv)
return _WeightedCVIterableWrapper(cv)
return cv # New style cv objects are passed without any modification