def test_suppress_warnings(self):
bad_log_dir = "path/to/bad_log/dir/zxc8aj3n"
with captured_output() as out:
self.prob.solve(self.prob.compute_max_l1_penalty(),
log_directory=bad_log_dir)
self.assertIn("Warning", out[0])
blitzml.suppress_warnings()
with captured_output() as out:
self.prob.solve(self.prob.compute_max_l1_penalty(),
log_directory=bad_log_dir)
self.assertNotIn("Warning", out[0])
blitzml.unsuppress_warnings()
def test_sparse_logreg_all_positive_labels_warning(self):
b = np.array([0., 1.0, 0.5])
A = np.zeros((3, 3))
with captured_output() as out:
prob = blitzml.SparseLogisticRegressionProblem(A, b)
message = out[0]
self.assertIn("Warning", message)
def test_sparse_logreg_bad_label_too_small(self):
b = np.array([-1., 0., -2.])
A = np.zeros((3, 3))
with captured_output() as out:
prob = blitzml.SparseLogisticRegressionProblem(A, b)
message = out[0]
self.assertIn("Warning", message)
def test_lasso_empty(self):
A = sp.csc_matrix((100, 1000))
b = np.ones(100)
with captured_output() as out:
prob = blitzml.LassoProblem(A, b)
self.assertEqual(prob.compute_max_l1_penalty(), 0.)
sol = prob.solve(0.)
self.assertEqual(np.linalg.norm(sol.weights), 0.)
self.assertEqual(sol.bias, 1.0)
self.assertEqual(sol.objective_value, 0.)
self.assertEqual(sol.duality_gap, 0.)
def setUp(self):
A = np.array([[2]], dtype=np.float)
b = np.array([10], dtype=np.double)
with captured_output() as out:
self.prob = blitzml.LassoProblem(A, b)
def test_lasso_csr_matrix_warning(self):
self.A = sp.csr_matrix(self.A)
with captured_output() as out:
prob = blitzml.LassoProblem(self.A, self.b)
message = out[0]
self.assertTrue("Warning" in message)
def test_lasso_dense_copy_warning_F_contiguous(self):
self.A = np.asfortranarray(self.A)
with captured_output() as out:
prob = blitzml.LassoProblem(self.A, self.b)
message = out[0]
self.assertTrue("Warning" not in message)