def test_obvious_clustering():
random = np.random.RandomState(0)
X = random.randn(40, 2)
X[20:] += 5
k1 = _MiniBatchKMedoids(n_clusters=2, max_iter=5, random_state=0, batch_size=20).fit(X)
k2 = _KMedoids(n_clusters=2, n_passes=10).fit(X)
assert (np.all(k1.labels_ == k2.labels_) or
np.all(k1.labels_ == np.logical_not(k2.labels_)))
def test_inertia():
random = np.random.RandomState(0)
X = random.randn(10, 2)
for km in (_KMedoids(n_clusters=3,
n_passes=5), _MiniBatchKMedoids(n_clusters=3)):
km.fit(X)
inertia = 0
for i in range(len(X)):
inertia += euclidean(X[km.cluster_ids_[km.labels_[i]]], X[i])
np.testing.assert_almost_equal(inertia, km.inertia_)
def test_inertia():
random = np.random.RandomState(0)
X = random.randn(10, 2)
for km in (_KMedoids(n_clusters=3, n_passes=5),
_MiniBatchKMedoids(n_clusters=3)):
km.fit(X)
inertia = 0
for i in range(len(X)):
inertia += euclidean(X[km.cluster_ids_[km.labels_[i]]], X[i])
np.testing.assert_almost_equal(inertia, km.inertia_)
def test_obvious_clustering():
random = np.random.RandomState(0)
X = random.randn(40, 2)
X[20:] += 5
k1 = _MiniBatchKMedoids(n_clusters=2,
max_iter=5,
random_state=0,
batch_size=20).fit(X)
k2 = _KMedoids(n_clusters=2, n_passes=10).fit(X)
assert (np.all(k1.labels_ == k2.labels_)
or np.all(k1.labels_ == np.logical_not(k2.labels_)))
def medoid():
return _KMedoids(metric='asdf').fit(np.zeros((10, 2)))
def test_4():
assert_raises(ValueError, lambda: _MiniBatchKMedoids(metric='sdf').fit(np.zeros((10,2))))
assert_raises(ValueError, lambda: _KMedoids(metric='sdf').fit(np.zeros((10,2))))
def test_4():
assert_raises(
ValueError,
lambda: _MiniBatchKMedoids(metric='sdf').fit(np.zeros((10, 2))))
assert_raises(ValueError,
lambda: _KMedoids(metric='sdf').fit(np.zeros((10, 2))))
def medoid():
return _KMedoids(metric='asdf').fit(np.zeros((10, 2)))