def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
"log": LogisticRegression(penalty="l1"),
"squared_hinge": LinearSVC(loss="squared_hinge", penalty="l1", dual=False),
}[loss]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) == 0).all())
assert_true((np.asarray(clf.intercept_) == 0).all())
clf.C = min_c * 1.01
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) != 0).any() or (np.asarray(clf.intercept_) != 0).any())
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
'log': LogisticRegression(penalty='l1'),
'l2': LinearSVC(loss='l2', penalty='l1', dual=False),
}[loss]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) == 0).all())
assert_true((np.asarray(clf.intercept_) == 0).all())
clf.C = min_c * 1.01
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) != 0).any() or
(np.asarray(clf.intercept_) != 0).any())
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
'log': LogisticRegression(penalty='l1'),
'l2': LinearSVC(loss='l2', penalty='l1', dual=False),
}[loss]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) == 0).all())
assert_true((np.asarray(clf.intercept_) == 0).all())
clf.C = min_c * 1.01
clf.fit(X, y)
assert_true((np.asarray(clf.coef_) != 0).any()
or (np.asarray(clf.intercept_) != 0).any())
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
'log': LogisticRegression(penalty='l1', solver='liblinear',
multi_class='ovr'),
'squared_hinge': LinearSVC(loss='squared_hinge',
penalty='l1', dual=False),
}[loss]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert (np.asarray(clf.coef_) == 0).all()
assert (np.asarray(clf.intercept_) == 0).all()
clf.C = min_c * 1.01
clf.fit(X, y)
assert ((np.asarray(clf.coef_) != 0).any() or
(np.asarray(clf.intercept_) != 0).any())
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
('log', False): LogisticRegression(penalty='l1'),
('log', True): SparseLogRegression(penalty='l1'),
('l2', False): LinearSVC(loss='l2', penalty='l1', dual=False),
('l2', True): SparseSVC(loss='l2', penalty='l1', dual=False),
}[loss, sp.issparse(X)]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert (np.asanyarray(clf.coef_) == 0).all()
assert (np.asanyarray(clf.intercept_) == 0).all()
clf.C = min_c * 1.01
clf.fit(X, y)
assert (np.asanyarray(clf.coef_) != 0).any() or \
(np.asanyarray(clf.intercept_) != 0).any()
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
('log', False): LogisticRegression(penalty='l1'),
('log', True): SparseLogRegression(penalty='l1'),
('l2', False): LinearSVC(loss='l2', penalty='l1', dual=False),
('l2', True): SparseSVC(loss='l2', penalty='l1', dual=False),
}[loss, sp.issparse(X)]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert (np.asarray(clf.coef_) == 0).all()
assert (np.asarray(clf.intercept_) == 0).all()
clf.C = min_c * 1.01
clf.fit(X, y)
assert (np.asarray(clf.coef_) != 0).any() or \
(np.asarray(clf.intercept_) != 0).any()
def check_l1_min_c(X, y, loss, fit_intercept=True, intercept_scaling=None):
min_c = l1_min_c(X, y, loss, fit_intercept, intercept_scaling)
clf = {
'log':
LogisticRegression(penalty='l1', solver='liblinear',
multi_class='ovr'),
'squared_hinge':
LinearSVC(loss='squared_hinge', penalty='l1', dual=False),
}[loss]
clf.fit_intercept = fit_intercept
clf.intercept_scaling = intercept_scaling
clf.C = min_c
clf.fit(X, y)
assert (np.asarray(clf.coef_) == 0).all()
assert (np.asarray(clf.intercept_) == 0).all()
clf.C = min_c * 1.01
clf.fit(X, y)
assert ((np.asarray(clf.coef_) != 0).any()
or (np.asarray(clf.intercept_) != 0).any())
def test_unsupported_loss():
l1_min_c(dense_X, Y1, 'l1')
def test_ill_posed_min_c():
X = [[0, 0], [0, 0]]
y = [0, 1]
l1_min_c(X, y)
def test_l2_deprecation():
clean_warning_registry()
with warnings.catch_warnings(record=True) as w:
assert_equal(l1_min_c(dense_X, Y1, "l2"),
l1_min_c(dense_X, Y1, "squared_hinge"))
assert_equal(w[0].category, DeprecationWarning)
def test_unsupported_loss():
l1_min_c(dense_X, Y1, 'l1')
def test_ill_posed_min_c():
X = [[0, 0], [0, 0]]
y = [0, 1]
l1_min_c(X, y)
def _check_param_grid(estimator, X, y, param_grid=None):
"""Check param_grid and return sensible default if param_grid is None.
Parameters
-----------
estimator: str, optional
The estimator to choose among: 'svc', 'svc_l2', 'svc_l1', 'logistic',
'logistic_l1', 'logistic_l2', 'ridge', 'ridge_classifier',
'ridge_regressor', and 'svr'. Note that the 'svc' and 'svc_l2';
'logistic' and 'logistic_l2'; 'ridge' and 'ridge_regressor'
correspond to the same estimator. Default 'svc'.
X: list of Niimg-like objects
See http://nilearn.github.io/manipulating_images/input_output.html
Data on which model is to be fitted. If this is a list,
the affine is considered the same for all.
y: array or list of shape (n_samples)
The dependent variable (age, sex, IQ, yes/no, etc.).
Target variable to predict. Must have exactly as many elements as
3D images in niimg.
param_grid: dict of str to sequence, or sequence of such. Default None
The parameter grid to explore, as a dictionary mapping estimator
parameters to sequences of allowed values.
An empty dict signifies default parameters.
A sequence of dicts signifies a sequence of grids to search, and is
useful to avoid exploring parameter combinations that make no sense
or have no effect. See scikit-learn documentation for more information.
Returns
-------
param_grid: dict of str to sequence, or sequence of such. Sensible default
dict has size 1.
"""
if param_grid is None:
param_grid = {}
# define loss function
if isinstance(estimator, LogisticRegression):
loss = 'log'
elif isinstance(estimator, (LinearSVC, _BaseRidgeCV, SVR)):
loss = 'squared_hinge'
else:
raise ValueError(
"Invalid estimator. The supported estimators are: {}".format(
list(SUPPORTED_ESTIMATORS.keys())))
# define sensible default for different types of estimators
if hasattr(estimator, 'penalty') and (estimator.penalty == 'l1'):
min_c = l1_min_c(X, y, loss=loss)
else:
min_c = 0.5
if not isinstance(estimator, _BaseRidgeCV):
param_grid['C'] = np.array([2, 20, 200]) * min_c
else:
param_grid = {}
return param_grid
def test_unsupported_loss():
with pytest.raises(ValueError):
l1_min_c(dense_X, Y1, 'l1')
def test_ill_posed_min_c():
X = [[0, 0], [0, 0]]
y = [0, 1]
with pytest.raises(ValueError):
l1_min_c(X, y)
def test_l2_deprecation():
clean_warning_registry()
with warnings.catch_warnings(record=True) as w:
assert_equal(l1_min_c(dense_X, Y1, "l2"),
l1_min_c(dense_X, Y1, "squared_hinge"))
assert_equal(w[0].category, DeprecationWarning)
