def test_enet_positive_constraint():
X = [[-1], [0], [1]]
y = [1, 0, -1] # just a straight line with negative slope
enet = ElasticNet(alpha=0.1, max_iter=1000, positive=True)
enet.fit(X, y)
assert min(enet.coef_) >= 0
def test_sparse_coef():
# Check that the sparse_coef property works
clf = ElasticNet()
clf.coef_ = [1, 2, 3]
assert sp.isspmatrix(clf.sparse_coef_)
assert clf.sparse_coef_.toarray().tolist()[0] == clf.coef_
def test_enet_float_precision():
# Generate dataset
X, y, X_test, y_test = build_dataset(n_samples=20, n_features=10)
# Here we have a small number of iterations, and thus the
# ElasticNet might not converge. This is to speed up tests
for normalize in [True, False]:
for fit_intercept in [True, False]:
coef = {}
intercept = {}
for dtype in [np.float64, np.float32]:
clf = ElasticNet(alpha=0.5,
max_iter=100,
precompute=False,
fit_intercept=fit_intercept,
normalize=normalize)
X = dtype(X)
y = dtype(y)
ignore_warnings(clf.fit)(X, y)
coef[('simple', dtype)] = clf.coef_
intercept[('simple', dtype)] = clf.intercept_
assert clf.coef_.dtype == dtype
# test precompute Gram array
Gram = X.T.dot(X)
clf_precompute = ElasticNet(alpha=0.5,
max_iter=100,
precompute=Gram,
fit_intercept=fit_intercept,
normalize=normalize)
ignore_warnings(clf_precompute.fit)(X, y)
assert_array_almost_equal(clf.coef_, clf_precompute.coef_)
assert_array_almost_equal(clf.intercept_,
clf_precompute.intercept_)
# test multi task enet
multi_y = np.hstack((y[:, np.newaxis], y[:, np.newaxis]))
clf_multioutput = MultiTaskElasticNet(
alpha=0.5,
max_iter=100,
fit_intercept=fit_intercept,
normalize=normalize)
clf_multioutput.fit(X, multi_y)
coef[('multi', dtype)] = clf_multioutput.coef_
intercept[('multi', dtype)] = clf_multioutput.intercept_
assert clf.coef_.dtype == dtype
for v in ['simple', 'multi']:
assert_array_almost_equal(coef[(v, np.float32)],
coef[(v, np.float64)],
decimal=4)
assert_array_almost_equal(intercept[(v, np.float32)],
intercept[(v, np.float64)],
decimal=4)
def test_enet_copy_X_True(check_input):
X, y, _, _ = build_dataset()
X = X.copy(order='F')
original_X = X.copy()
enet = ElasticNet(copy_X=True)
enet.fit(X, y, check_input=check_input)
assert_array_equal(original_X, X)
def test_warm_start():
X, y, _, _ = build_dataset()
clf = ElasticNet(alpha=0.1, max_iter=5, warm_start=True)
ignore_warnings(clf.fit)(X, y)
ignore_warnings(clf.fit)(X, y) # do a second round with 5 iterations
clf2 = ElasticNet(alpha=0.1, max_iter=10)
ignore_warnings(clf2.fit)(X, y)
assert_array_almost_equal(clf2.coef_, clf.coef_)
def test_enet_copy_X_False_check_input_False():
X, y, _, _ = build_dataset()
X = X.copy(order='F')
original_X = X.copy()
enet = ElasticNet(copy_X=False)
enet.fit(X, y, check_input=False)
# No copying, X is overwritten
assert np.any(np.not_equal(original_X, X))
def test_overrided_gram_matrix():
X, y, _, _ = build_dataset(n_samples=20, n_features=10)
Gram = X.T.dot(X)
clf = ElasticNet(selection='cyclic', tol=1e-8, precompute=Gram)
assert_warns_message(
UserWarning, "Gram matrix was provided but X was centered"
" to fit intercept, "
"or X was normalized : recomputing Gram matrix.", clf.fit, X, y)
def test_normalize_option():
# Check that the normalize option in enet works
X = sp.csc_matrix([[-1], [0], [1]])
y = [-1, 0, 1]
clf_dense = ElasticNet(normalize=True)
clf_sparse = ElasticNet(normalize=True)
clf_dense.fit(X, y)
X = sp.csc_matrix(X)
clf_sparse.fit(X, y)
assert_almost_equal(clf_dense.dual_gap_, 0)
assert_array_almost_equal(clf_dense.coef_, clf_sparse.coef_)
def _test_sparse_enet_not_as_toy_dataset(alpha, fit_intercept, positive):
n_samples, n_features, max_iter = 100, 100, 1000
n_informative = 10
X, y = make_sparse_data(n_samples,
n_features,
n_informative,
positive=positive)
X_train, X_test = X[n_samples // 2:], X[:n_samples // 2]
y_train, y_test = y[n_samples // 2:], y[:n_samples // 2]
s_clf = ElasticNet(alpha=alpha,
l1_ratio=0.8,
fit_intercept=fit_intercept,
max_iter=max_iter,
tol=1e-7,
positive=positive,
warm_start=True)
s_clf.fit(X_train, y_train)
assert_almost_equal(s_clf.dual_gap_, 0, 4)
assert s_clf.score(X_test, y_test) > 0.85
# check the convergence is the same as the dense version
d_clf = ElasticNet(alpha=alpha,
l1_ratio=0.8,
fit_intercept=fit_intercept,
max_iter=max_iter,
tol=1e-7,
positive=positive,
warm_start=True)
d_clf.fit(X_train.toarray(), y_train)
assert_almost_equal(d_clf.dual_gap_, 0, 4)
assert d_clf.score(X_test, y_test) > 0.85
assert_almost_equal(s_clf.coef_, d_clf.coef_, 5)
assert_almost_equal(s_clf.intercept_, d_clf.intercept_, 5)
# check that the coefs are sparse
assert np.sum(s_clf.coef_ != 0.0) < 2 * n_informative
def test_warm_start_convergence_with_regularizer_decrement():
X, y = load_boston(return_X_y=True)
# Train a model to converge on a lightly regularized problem
final_alpha = 1e-5
low_reg_model = ElasticNet(alpha=final_alpha).fit(X, y)
# Fitting a new model on a more regularized version of the same problem.
# Fitting with high regularization is easier it should converge faster
# in general.
high_reg_model = ElasticNet(alpha=final_alpha * 10).fit(X, y)
assert low_reg_model.n_iter_ > high_reg_model.n_iter_
# Fit the solution to the original, less regularized version of the
# problem but from the solution of the highly regularized variant of
# the problem as a better starting point. This should also converge
# faster than the original model that starts from zero.
warm_low_reg_model = deepcopy(high_reg_model)
warm_low_reg_model.set_params(warm_start=True, alpha=final_alpha)
warm_low_reg_model.fit(X, y)
assert low_reg_model.n_iter_ > warm_low_reg_model.n_iter_
def test_sparse_input_convergence_warning():
X, y, _, _ = build_dataset(n_samples=1000, n_features=500)
with pytest.warns(ConvergenceWarning):
ElasticNet(max_iter=1,
tol=0).fit(sparse.csr_matrix(X, dtype=np.float32), y)
# check that the model converges w/o warnings
with pytest.warns(None) as record:
Lasso(max_iter=1000).fit(sparse.csr_matrix(X, dtype=np.float32), y)
assert not record.list
def test_precompute_invalid_argument():
X, y, _, _ = build_dataset()
for clf in [
ElasticNetCV(precompute="invalid"),
LassoCV(precompute="invalid")
]:
assert_raises_regex(
ValueError, ".*should be.*True.*False.*auto.*"
"array-like.*Got 'invalid'", clf.fit, X, y)
# Precompute = 'auto' is not supported for ElasticNet
assert_raises_regex(ValueError, ".*should be.*True.*False.*array-like.*"
"Got 'auto'",
ElasticNet(precompute='auto').fit, X, y)
def test_warm_start_convergence():
X, y, _, _ = build_dataset()
model = ElasticNet(alpha=1e-3, tol=1e-3).fit(X, y)
n_iter_reference = model.n_iter_
# This dataset is not trivial enough for the model to converge in one pass.
assert n_iter_reference > 2
# Check that n_iter_ is invariant to multiple calls to fit
# when warm_start=False, all else being equal.
model.fit(X, y)
n_iter_cold_start = model.n_iter_
assert n_iter_cold_start == n_iter_reference
# Fit the same model again, using a warm start: the optimizer just performs
# a single pass before checking that it has already converged
model.set_params(warm_start=True)
model.fit(X, y)
n_iter_warm_start = model.n_iter_
assert n_iter_warm_start == 1
def test_enet_multitarget():
n_targets = 3
X, y = make_sparse_data(n_targets=n_targets)
estimator = ElasticNet(alpha=0.01, precompute=None)
# XXX: There is a bug when precompute is not None!
estimator.fit(X, y)
coef, intercept, dual_gap = (estimator.coef_, estimator.intercept_,
estimator.dual_gap_)
for k in range(n_targets):
estimator.fit(X, y[:, k])
assert_array_almost_equal(coef[k, :], estimator.coef_)
assert_array_almost_equal(intercept[k], estimator.intercept_)
assert_array_almost_equal(dual_gap[k], estimator.dual_gap_)
def test_enet_multitarget():
n_targets = 3
X, y, _, _ = build_dataset(n_samples=10,
n_features=8,
n_informative_features=10,
n_targets=n_targets)
estimator = ElasticNet(alpha=0.01)
estimator.fit(X, y)
coef, intercept, dual_gap = (estimator.coef_, estimator.intercept_,
estimator.dual_gap_)
for k in range(n_targets):
estimator.fit(X, y[:, k])
assert_array_almost_equal(coef[k, :], estimator.coef_)
assert_array_almost_equal(intercept[k], estimator.intercept_)
assert_array_almost_equal(dual_gap[k], estimator.dual_gap_)
def test_check_input_false():
X, y, _, _ = build_dataset(n_samples=20, n_features=10)
X = check_array(X, order='F', dtype='float64')
y = check_array(X, order='F', dtype='float64')
clf = ElasticNet(selection='cyclic', tol=1e-8)
# Check that no error is raised if data is provided in the right format
clf.fit(X, y, check_input=False)
# With check_input=False, an exhaustive check is not made on y but its
# dtype is still cast in _preprocess_data to X's dtype. So the test should
# pass anyway
X = check_array(X, order='F', dtype='float32')
clf.fit(X, y, check_input=False)
# With no input checking, providing X in C order should result in false
# computation
X = check_array(X, order='C', dtype='float64')
assert_raises(ValueError, clf.fit, X, y, check_input=False)
def test_same_multiple_output_sparse_dense():
for normalize in [True, False]:
l = ElasticNet(normalize=normalize)
X = [[0, 1, 2, 3, 4], [0, 2, 5, 8, 11], [9, 10, 11, 12, 13],
[10, 11, 12, 13, 14]]
y = [[1, 2, 3, 4, 5], [1, 3, 6, 9, 12], [10, 11, 12, 13, 14],
[11, 12, 13, 14, 15]]
ignore_warnings(l.fit)(X, y)
sample = np.array([1, 2, 3, 4, 5]).reshape(1, -1)
predict_dense = l.predict(sample)
l_sp = ElasticNet(normalize=normalize)
X_sp = sp.coo_matrix(X)
ignore_warnings(l_sp.fit)(X_sp, y)
sample_sparse = sp.coo_matrix(sample)
predict_sparse = l_sp.predict(sample_sparse)
assert_array_almost_equal(predict_sparse, predict_dense)
def test_enet_toy_explicit_sparse_input():
# Test ElasticNet for various values of alpha and l1_ratio with sparse X
f = ignore_warnings
# training samples
X = sp.lil_matrix((3, 1))
X[0, 0] = -1
# X[1, 0] = 0
X[2, 0] = 1
Y = [-1, 0, 1] # just a straight line (the identity function)
# test samples
T = sp.lil_matrix((3, 1))
T[0, 0] = 2
T[1, 0] = 3
T[2, 0] = 4
# this should be the same as lasso
clf = ElasticNet(alpha=0, l1_ratio=1.0)
f(clf.fit)(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [1])
assert_array_almost_equal(pred, [2, 3, 4])
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.3, max_iter=1000)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
assert_array_almost_equal(pred, [1.0163, 1.5245, 2.0327], decimal=3)
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.5)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.45454], 3)
assert_array_almost_equal(pred, [0.9090, 1.3636, 1.8181], 3)
assert_almost_equal(clf.dual_gap_, 0)
def test_enet_toy_list_input():
# Test ElasticNet for various values of alpha and l1_ratio with list X
X = np.array([[-1], [0], [1]])
X = sp.csc_matrix(X)
Y = [-1, 0, 1] # just a straight line
T = np.array([[2], [3], [4]]) # test sample
# this should be the same as unregularized least squares
clf = ElasticNet(alpha=0, l1_ratio=1.0)
# catch warning about alpha=0.
# this is discouraged but should work.
ignore_warnings(clf.fit)(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [1])
assert_array_almost_equal(pred, [2, 3, 4])
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.3, max_iter=1000)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
assert_array_almost_equal(pred, [1.0163, 1.5245, 2.0327], decimal=3)
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.5)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.45454], 3)
assert_array_almost_equal(pred, [0.9090, 1.3636, 1.8181], 3)
assert_almost_equal(clf.dual_gap_, 0)
def test_enet_toy():
# Test ElasticNet for various parameters of alpha and l1_ratio.
# Actually, the parameters alpha = 0 should not be allowed. However,
# we test it as a border case.
# ElasticNet is tested with and without precomputed Gram matrix
X = np.array([[-1.], [0.], [1.]])
Y = [-1, 0, 1] # just a straight line
T = [[2.], [3.], [4.]] # test sample
# this should be the same as lasso
clf = ElasticNet(alpha=1e-8, l1_ratio=1.0)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [1])
assert_array_almost_equal(pred, [2, 3, 4])
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.3, max_iter=100, precompute=False)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
assert_array_almost_equal(pred, [1.0163, 1.5245, 2.0327], decimal=3)
assert_almost_equal(clf.dual_gap_, 0)
clf.set_params(max_iter=100, precompute=True)
clf.fit(X, Y) # with Gram
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
assert_array_almost_equal(pred, [1.0163, 1.5245, 2.0327], decimal=3)
assert_almost_equal(clf.dual_gap_, 0)
clf.set_params(max_iter=100, precompute=np.dot(X.T, X))
clf.fit(X, Y) # with Gram
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.50819], decimal=3)
assert_array_almost_equal(pred, [1.0163, 1.5245, 2.0327], decimal=3)
assert_almost_equal(clf.dual_gap_, 0)
clf = ElasticNet(alpha=0.5, l1_ratio=0.5)
clf.fit(X, Y)
pred = clf.predict(T)
assert_array_almost_equal(clf.coef_, [0.45454], 3)
assert_array_almost_equal(pred, [0.9090, 1.3636, 1.8181], 3)
assert_almost_equal(clf.dual_gap_, 0)
def test_random_descent():
# Test that both random and cyclic selection give the same results.
# Ensure that the test models fully converge and check a wide
# range of conditions.
# This uses the coordinate descent algo using the gram trick.
X, y, _, _ = build_dataset(n_samples=50, n_features=20)
clf_cyclic = ElasticNet(selection='cyclic', tol=1e-8)
clf_cyclic.fit(X, y)
clf_random = ElasticNet(selection='random', tol=1e-8, random_state=42)
clf_random.fit(X, y)
assert_array_almost_equal(clf_cyclic.coef_, clf_random.coef_)
assert_almost_equal(clf_cyclic.intercept_, clf_random.intercept_)
# This uses the descent algo without the gram trick
clf_cyclic = ElasticNet(selection='cyclic', tol=1e-8)
clf_cyclic.fit(X.T, y[:20])
clf_random = ElasticNet(selection='random', tol=1e-8, random_state=42)
clf_random.fit(X.T, y[:20])
assert_array_almost_equal(clf_cyclic.coef_, clf_random.coef_)
assert_almost_equal(clf_cyclic.intercept_, clf_random.intercept_)
# Sparse Case
clf_cyclic = ElasticNet(selection='cyclic', tol=1e-8)
clf_cyclic.fit(sparse.csr_matrix(X), y)
clf_random = ElasticNet(selection='random', tol=1e-8, random_state=42)
clf_random.fit(sparse.csr_matrix(X), y)
assert_array_almost_equal(clf_cyclic.coef_, clf_random.coef_)
assert_almost_equal(clf_cyclic.intercept_, clf_random.intercept_)
# Multioutput case.
new_y = np.hstack((y[:, np.newaxis], y[:, np.newaxis]))
clf_cyclic = MultiTaskElasticNet(selection='cyclic', tol=1e-8)
clf_cyclic.fit(X, new_y)
clf_random = MultiTaskElasticNet(selection='random',
tol=1e-8,
random_state=42)
clf_random.fit(X, new_y)
assert_array_almost_equal(clf_cyclic.coef_, clf_random.coef_)
assert_almost_equal(clf_cyclic.intercept_, clf_random.intercept_)
# Raise error when selection is not in cyclic or random.
clf_random = ElasticNet(selection='invalid')
assert_raises(ValueError, clf_random.fit, X, y)
