def do_test_kmeans_results(self, representation, algo, dtype):
# cheks that kmeans works as intended
array_constr = {
'dense': np.array,
'sparse': sp.csr_matrix
}[representation]
X = array_constr([[0, 0], [0.5, 0], [0.5, 1], [1, 1]], dtype=dtype)
# will be rescaled to [1.5, 0.5, 0.5, 1.5]
sample_weight = [3, 1, 1, 3]
init_centers = np.array([[0, 0], [1, 1]], dtype=dtype)
expected_labels = [0, 0, 1, 1]
expected_inertia = 0.1875
expected_centers = np.array([[0.125, 0], [0.875, 1]], dtype=dtype)
expected_n_iter = 2
kmeans = KMeansL1L2(n_clusters=2,
n_init=1,
init=init_centers,
algorithm=algo)
kmeans.fit(X, sample_weight=sample_weight)
assert_array_equal(kmeans.labels_, expected_labels)
assert_almost_equal(kmeans.inertia_, expected_inertia)
assert_array_almost_equal(kmeans.cluster_centers_, expected_centers)
self.assertEqualArray(kmeans.n_iter_, expected_n_iter)
def test_kmeans_l2_iris(self):
iris = datasets.load_iris()
X = iris.data
clr = KMeansL1L2(4, norm='L2')
clr.fit(X)
cls = set(clr.predict(X))
self.assertEqual({0, 1, 2, 3}, cls)
def test_k_means_random_init_not_precomputed(self):
km = KMeansL1L2(init="random",
n_clusters=TestKMeansL1L2Sklearn.n_clusters,
random_state=42,
precompute_distances=False).fit(
TestKMeansL1L2Sklearn.X)
self._check_fitted_model(km)
def test_kmeans_l2_random(self):
iris = datasets.load_iris()
X = iris.data
clr = KMeansL1L2(4, init="random")
clr.fit(X)
cls = set(clr.predict(X))
self.assertEqual({0, 1, 2, 3}, cls)
def test_kmeans_l1_small(self):
iris = datasets.load_iris()
X = iris.data
X = X[:6]
clr = KMeansL1L2(4, norm='L1', n_jobs=1)
clr.fit(X)
cls = set(clr.predict(X))
self.assertEqual({0, 1, 2, 3}, cls)
def do_test_kmeans_results(self, representation, algo, dtype, norm, sw):
# cheks that kmeans works as intended
array_constr = {
'dense': np.array,
'sparse': sp.csr_matrix
}[representation]
X = array_constr([[0, 0], [0.5, 0], [0.5, 1], [1, 1]], dtype=dtype)
init_centers = np.array([[0, 0], [1, 1]], dtype=dtype)
# will be rescaled to [1.5, 0.5, 0.5, 1.5]
if sw:
sample_weight = [3, 1, 1, 3]
if sklearn_023:
expected_inertia = 0.375
else:
expected_inertia = 0.1875
expected_centers = np.array([[0.125, 0], [0.875, 1]], dtype=dtype)
expected_n_iter = 2
else:
sample_weight = None
if norm == 'L2':
expected_inertia = 0.25
expected_centers = np.array([[0.25, 0], [0.75, 1]],
dtype=dtype)
expected_n_iter = 2
else:
expected_inertia = 1.
expected_centers = np.array([[0.25, 0], [0.75, 1]],
dtype=dtype)
expected_n_iter = 1
expected_labels = [0, 0, 1, 1]
try:
kmeans = KMeansL1L2(n_clusters=2,
n_init=1,
init=init_centers,
algorithm=algo,
norm=norm)
except NotImplementedError as e:
if ("Only algorithm 'full' is implemented" in str(e)
and norm == 'L1'):
return
raise e
try:
kmeans.fit(X, sample_weight=sample_weight)
except NotImplementedError as e:
if ("Non uniform weights are not implemented yet" in str(e)
and norm == 'L1'):
return
if ("Sparse matrix is not implemented" in str(e) and norm == 'L1'):
return
raise e
assert_array_equal(kmeans.labels_, expected_labels)
assert_almost_equal(kmeans.inertia_, expected_inertia)
assert_array_almost_equal(kmeans.cluster_centers_, expected_centers)
self.assertEqualArray(kmeans.n_iter_, expected_n_iter)
def test_kmeans_l1_check(self):
X = numpy.array([[-10, 1, 2, 3, 4, 10], [-10, 1, 2, 3, 4, 10]]).T
clr = KMeansL1L2(2, norm='L1')
clr.fit(X)
cls = set(clr.predict(X))
self.assertEqual({0, 1}, cls)
self.assertEqual(clr.cluster_centers_.shape, (2, 2))
self.assertEqualArray(clr.cluster_centers_.max(), [3, 3])
tr = clr.transform(X)
self.assertEqual(tr.shape, (X.shape[0], 2))
tr = clr.transform([[3, 3]])
self.assertEqualArray(tr.min(), [0])
def test_k_means_new_centers(self):
# Explore the part of the code where a new center is reassigned
X = np.array([[0, 0, 1, 1], [0, 0, 0, 0], [0, 1, 0, 0], [0, 0, 0, 0],
[0, 0, 0, 0], [0, 1, 0, 0]])
labels = [0, 1, 2, 1, 1, 2]
bad_centers = np.array([[+0, 1, 0, 0], [.2, 0, .2, .2], [+0, 0, 0, 0]])
km = KMeansL1L2(n_clusters=3,
init=bad_centers,
n_init=1,
max_iter=10,
random_state=1)
for this_X in (X, sp.coo_matrix(X)):
km.fit(this_X)
this_labels = km.labels_
# Reorder the labels so that the first instance is in cluster 0,
# the second in cluster 1, ...
this_labels = np.unique(this_labels,
return_index=True)[1][this_labels]
np.testing.assert_array_equal(this_labels, labels)
def test_k_means_plus_plus_init_not_precomputed(self):
km = KMeansL1L2(init="k-means++",
n_clusters=TestKMeansL1L2Sklearn.n_clusters,
random_state=42).fit(TestKMeansL1L2Sklearn.X)
self._check_fitted_model(km)