TST add unit testing for inf weights (#234)

Co-authored-by: mathurinm <[email protected]>

celer/tests/test_lasso.py

@@ -8,7 +8,7 @@
 
 import numpy as np
 from numpy.linalg import norm
-from numpy.testing import assert_allclose, assert_array_less
+from numpy.testing import assert_allclose, assert_array_less, assert_array_equal
 import pytest
 
 from sklearn.exceptions import ConvergenceWarning
@@ -218,6 +218,33 @@ def test_weights_lasso():
     np.testing.assert_raises(ValueError, clf1.fit, X=X, y=y)
 
 
+@pytest.mark.parametrize("pb", ["lasso", "logreg"])
+def test_infinite_weights(pb):
+    n_samples, n_features = 50, 100
+    X, y = build_dataset(n_samples, n_features)
+    if pb == "logreg":
+        y = np.sign(y)
+
+    np.random.seed(1)
+    weights = np.abs(np.random.randn(n_features))
+    n_inf = n_features // 10
+    inf_indices = np.random.choice(n_features, size=n_inf, replace=False)
+    weights[inf_indices] = np.inf
+
+    alpha = norm(X.T @ y / weights, ord=np.inf) / n_samples / 100
+
+    tol = 1e-8
+    _, coefs, dual_gaps = celer_path(
+        X, y, pb=pb, alphas=[alpha], weights=weights, tol=tol)
+
+    if pb == "logreg":
+        assert_array_less(dual_gaps[0], tol * n_samples * np.log(2))
+    else:
+        assert_array_less(dual_gaps[0], tol * norm(y) ** 2 / 2.)
+
+    assert_array_equal(coefs[inf_indices], 0)
+
+
 def test_zero_iter():
     X, y = build_dataset(n_samples=30, n_features=50)
 

celer/tests/test_logreg.py

@@ -1,11 +1,10 @@
 # Author: Mathurin Massias <[email protected]>
 # License: BSD 3 clause
-
 import pytest
 import numpy as np
 from numpy.linalg import norm
 
-from numpy.testing import assert_allclose, assert_array_less
+from numpy.testing import assert_allclose, assert_array_less, assert_array_equal
 from sklearn.linear_model._logistic import _logistic_regression_path
 from sklearn.utils.estimator_checks import check_estimator
 from sklearn.linear_model import LogisticRegression as sklearn_Logreg

celer/tests/test_mtl.py

@@ -1,8 +1,10 @@
 import pytest
+import warnings
 import itertools
+
 import numpy as np
 from numpy.linalg import norm
-from numpy.testing import assert_allclose, assert_array_less
+from numpy.testing import assert_allclose, assert_array_less, assert_array_equal
 
 from sklearn.utils.estimator_checks import check_estimator
 from sklearn.linear_model import MultiTaskLassoCV as sklearn_MultiTaskLassoCV
@@ -241,5 +243,29 @@ def test_check_weights():
     np.testing.assert_raises(ValueError, clf.fit, X=X, y=y)
 
 
+def test_infinite_weights_group():
+    n_samples, n_features = 50, 100
+    X, y = build_dataset(n_samples, n_features)
+
+    np.random.seed(1)
+    group_size = 5
+    weights = np.abs(np.random.randn(n_features // group_size))
+    n_inf = 3
+    inf_indices = np.random.choice(
+        n_features // group_size, size=n_inf, replace=False)
+    weights[inf_indices] = np.inf
+    alpha_max = np.max(
+        norm((X.T @ y).reshape(-1, group_size), 2, axis=1)
+    ) / n_samples
+
+    clf = GroupLasso(
+        alpha=alpha_max / 100., weights=weights, groups=group_size, tol=1e-8
+    ).fit(X, y)
+
+    assert_array_less(clf.dual_gap_, clf.tol * norm(y) ** 2 / 2)
+    assert_array_equal(
+        norm(clf.coef_.reshape(-1, group_size), axis=1)[inf_indices], 0)
+
+
 if __name__ == "__main__":
     pass