def test_constrained_kmeans_relocated_clusters():
# check that empty clusters are relocated as expected
X = np.array([[0, 0], [0.5, 0], [0.5, 1], [1, 1]])
y = [-1, -1, -1, -1]
y_rejected = [set() for _ in range(len(y))]
# second center too far from others points will be empty at first iter
init_centers = np.array([[0.4, 0.4], [3, 3]])
r_mat, labels = rejection_matrix(y, y_rejected, n_classes=2)
# No given labels
assert labels == []
kmeans = ConstrainedKMeans(n_init=1, init=init_centers, verbose=True)
kmeans.fit(X, r_mat)
expected_labels = [0, 0, 1, 1]
expected_inertia = 0.25
expected_centers = [[0.25, 0], [0.75, 1]]
expected_n_iter = 3
expected_sample_distances = [0.0625, 0.0625, 0.0625, 0.0625]
assert_array_equal(kmeans.labels_, expected_labels)
assert_allclose(kmeans.inertia_, expected_inertia)
assert_allclose(kmeans.cluster_centers_, expected_centers)
assert_allclose(kmeans.sample_distances_, expected_sample_distances)
assert kmeans.n_iter_ == expected_n_iter
def test_rejection_matrix():
y = [-1, -1, -1, 0, 0, 0, 0, 1, 1, 1, 1]
y_rejected = [{0}, {0}, {0},
set(),
set(),
set(),
set(),
set(),
set(),
set(),
set()]
r_mat, labels = rejection_matrix(y, y_rejected)
assert labels == [0, 1]
assert_array_equal(
todense(r_mat),
[
[True, False],
[True, False],
[True, False],
[
False,
True,
],
[False, True],
[False, True],
[False, True],
[True, False],
[True, False],
[True, False],
[True, False],
],
)
def test_constrained_kmeans_label_persistence():
X, y = make_blobs()
# Non-continous labels shouldn't be a problem
y = 10 + 2 * y
# Reset 80% labels to -1
reset_indices = np.random.choice(y.shape[0], round(y.shape[0] * 0.8))
y[reset_indices] = -1
r_mat, labels = rejection_matrix(y)
labels = np.array(labels)
kmeans = ConstrainedKMeans(n_init=1, verbose=True)
kmeans.fit(X, r_mat)
mask_known = y > -1
assert_array_equal(labels[kmeans.labels_[mask_known]], y[mask_known])
def test_rejection_matrix_too_few_classes():
y = [1, 2, 3, 4, 5, 6, 7]
y_rejected = [set() for _ in range(len(y))]
with pytest.raises(ValueError):
rejection_matrix(y, y_rejected, n_classes=2)
def test_rejection_matrix_reject_all():
y = [-1, -1, -1, 0, 0, 0, 0]
y_rejected = [{0}, {0}, {0}, set(), set(), set(), set()]
with pytest.raises(RuntimeError, match="All classes were rejected"):
rejection_matrix(y, y_rejected)