def test_1d_flat_fit_and_predict():
X1 = np.arange(0, 6, 1).reshape(6, 1)
ms.flat = Mock(wraps=ms.flat)
clusters = ms.MeanShift(bandwidth=2.5)
centres1, labels1 = clusters.fit(X1)
ms.flat.assert_called()
assert centres1 == np.array([2.5])
assert labels1 == np.array([0])
assert clusters.predict_point(np.array(3)) == 0
assert clusters.predict_point(np.array(14)) == 0
X2 = np.hstack([np.arange(0, 3, 1), np.arange(6, 9, 1)]).reshape(6, 1)
clusters = ms.MeanShift(bandwidth=2, kernel='flat')
centres2, labels2 = clusters.fit(X2)
assert np.all(centres2 == np.array([1, 7]).reshape(2, 1))
assert np.all(labels2 == np.array([0, 1]))
assert clusters.predict_point(np.array(3)) == 0
assert clusters.predict_point(np.array(5)) == 1
def test_1d_gaussian_fit_and_predict():
X1 = np.arange(0, 6, 1).reshape(6, 1)
ms.gaussian = Mock(wraps=ms.gaussian)
clusters = ms.MeanShift(bandwidth=2, kernel='gaussian')
centres1, labels1 = clusters.fit(X1)
ms.gaussian.assert_called()
assert centres1 == np.array([2.5])
assert labels1 == np.array([0])
X2 = np.hstack([np.arange(0, 3, 1), np.arange(6, 9, 1)]).reshape(6, 1)
clusters2 = ms.MeanShift(bandwidth=2, kernel='gaussian')
centres2, labels2 = clusters2.fit(X2)
assert np.all(centres2 == pytest.approx(np.array([1, 7]).reshape(2, 1)))
assert np.all(labels2 == np.array([0, 1]))
to_predict = np.arange(3, 7).reshape(4, 1)
assert np.all(clusters.predict(to_predict) == 0)
exp = np.array([0, 0, 1, 1])
assert np.all(clusters2.predict(to_predict) == exp)
def test_2d_flat_fit_and_predict():
X = np.arange(0, 60).reshape(30, 2)
ms.flat = Mock(wraps=ms.flat)
clusters = ms.MeanShift(bandwidth=20)
centres, labels = clusters.fit(X)
ms.flat.assert_called()
exp = np.arange(29., 31.).reshape(1, 2)
assert np.all(centres == pytest.approx(exp))
assert labels == np.arange(0, 1)
clusters2 = ms.MeanShift(bandwidth=14, kernel='flat')
centres, labels = clusters2.fit(X)
c1 = np.arange(14.11844117, 16)
c2 = np.arange(29., 31.)
c3 = np.arange(43.88155883, 45)
exp = np.array([c1, c2, c3])
assert np.all(centres == pytest.approx(exp))
assert np.all(labels == np.arange(0, 3))
to_predict = np.array([[11, 15], [24, 34], [37, 39]])
assert np.all(clusters.predict(to_predict) == 0)
assert np.all(clusters2.predict(to_predict) == np.array([0, 1, 2]))
def test_2d_gaussian_fit_and_predict():
X = np.arange(0, 60).reshape(30, 2)
ms.gaussian = Mock(wraps=ms.gaussian)
clusters = ms.MeanShift(bandwidth=23, kernel='gaussian')
centres, labels = clusters.fit(X)
ms.gaussian.assert_called()
exp = np.arange(29., 31.).reshape(1, 2)
assert np.all(centres == pytest.approx(exp))
assert labels == np.arange(0, 1)
clusters2 = ms.MeanShift(bandwidth=14, kernel='gaussian')
centres, labels = clusters2.fit(X)
c1 = np.arange(13.60007775, 15)
c2 = np.arange(29., 31.)
c3 = np.arange(44.39992225, 46)
exp = np.array([c1, c2, c3])
assert np.all(centres == pytest.approx(exp))
assert np.all(labels == np.arange(0, 3))
to_predict = np.array([[11, 15], [24, 34], [37, 39]])
assert np.all(clusters.predict(to_predict) == 0)
assert np.all(clusters2.predict(to_predict) == np.array([0, 1, 2]))
def test_nd_flat_fit_and_predict():
X = np.arange(0, 120).reshape(30, 2, 2)
c = np.arange(58, 62).reshape(2, 2)
ms.flat = Mock(wraps=ms.flat)
clusters = ms.MeanShift(bandwidth=70)
centres, labels = clusters.fit(X)
ms.flat.assert_called()
assert np.all(centres == c)
assert labels == np.arange(0, 1)
clusters2 = ms.MeanShift(bandwidth=43, kernel='flat')
centres, labels = clusters2.fit(X)
c1 = np.arange(35.43210209, 39).reshape(2, 2)
c2 = np.arange(80.56789791, 84).reshape(2, 2)
exp = np.array([c1, c2])
assert np.all(centres == pytest.approx(exp))
assert np.all(labels == np.arange(0, 2))
to_predict = np.array([[[56, 56], [56, 56]], [[60, 60], [60, 60]]])
assert np.all(clusters.predict(to_predict) == 0)
assert np.all(clusters2.predict(to_predict) == np.array([0, 1]))
def test_invalid_kernel():
with pytest.raises(ValueError):
_ = ms.MeanShift(bandwidth=23, kernel='test')