def testNearestNeighbors(self):
rs = np.random.RandomState(0)
raw_X = rs.rand(10, 5)
raw_Y = rs.rand(8, 5)
X = mt.tensor(raw_X)
Y = mt.tensor(raw_Y)
raw_sparse_x = sps.random(10,
5,
density=0.5,
format='csr',
random_state=rs)
raw_sparse_y = sps.random(8,
5,
density=0.4,
format='csr',
random_state=rs)
X_sparse = mt.tensor(raw_sparse_x)
Y_sparse = mt.tensor(raw_sparse_y)
metric_func = lambda u, v: np.sqrt(((u - v)**2).sum())
_ = NearestNeighbors(algorithm='auto',
metric='precomputed',
metric_params={})
with self.assertRaises(ValueError):
_ = NearestNeighbors(algorithm='unknown')
with self.assertRaises(ValueError):
_ = NearestNeighbors(algorithm='kd_tree', metric=metric_func)
with self.assertRaises(ValueError):
_ = NearestNeighbors(algorithm='auto', metric='unknown')
assert_warns(SyntaxWarning, NearestNeighbors, metric_params={'p': 1})
with self.assertRaises(ValueError):
_ = NearestNeighbors(metric='wminkowski', p=0)
with self.assertRaises(ValueError):
_ = NearestNeighbors(algorithm='auto', metric='minkowski', p=0)
nn = NearestNeighbors(algorithm='auto', metric='minkowski', p=1)
nn.fit(X)
self.assertEqual(nn.effective_metric_, 'manhattan')
nn = NearestNeighbors(algorithm='auto', metric='minkowski', p=2)
nn.fit(X)
self.assertEqual(nn.effective_metric_, 'euclidean')
nn = NearestNeighbors(algorithm='auto', metric='minkowski', p=np.inf)
nn.fit(X)
self.assertEqual(nn.effective_metric_, 'chebyshev')
nn2 = NearestNeighbors(algorithm='auto', metric='minkowski')
nn2.fit(nn)
self.assertEqual(nn2._fit_method, nn._fit_method)
nn = NearestNeighbors(algorithm='auto', metric='minkowski')
ball_tree = SkBallTree(raw_X)
nn.fit(ball_tree)
self.assertEqual(nn._fit_method, 'ball_tree')
nn = NearestNeighbors(algorithm='auto', metric='minkowski')
kd_tree = SkKDTree(raw_X)
nn.fit(kd_tree)
self.assertEqual(nn._fit_method, 'kd_tree')
with self.assertRaises(ValueError):
nn = NearestNeighbors()
nn.fit(np.random.rand(0, 10))
nn = NearestNeighbors(algorithm='ball_tree')
assert_warns(UserWarning, nn.fit, X_sparse)
nn = NearestNeighbors(metric='haversine')
with self.assertRaises(ValueError):
nn.fit(X_sparse)
nn = NearestNeighbors(metric=metric_func, n_neighbors=1)
nn.fit(X)
self.assertEqual(nn._fit_method, 'ball_tree')
nn = NearestNeighbors(metric='sqeuclidean', n_neighbors=1)
nn.fit(X)
self.assertEqual(nn._fit_method, 'brute')
with self.assertRaises(ValueError):
nn = NearestNeighbors(n_neighbors=-1)
nn.fit(X)
with self.assertRaises(TypeError):
nn = NearestNeighbors(n_neighbors=1.3)
nn.fit(X)
nn = NearestNeighbors()
nn.fit(X)
with self.assertRaises(ValueError):
nn.kneighbors(Y, n_neighbors=-1)
with self.assertRaises(TypeError):
nn.kneighbors(Y, n_neighbors=1.3)
with self.assertRaises(ValueError):
nn.kneighbors(Y, n_neighbors=11)
nn = NearestNeighbors(algorithm='ball_tree')
nn.fit(X)
with self.assertRaises(ValueError):
nn.kneighbors(Y_sparse)
def test_nearest_neighbors(setup):
rs = np.random.RandomState(0)
raw_X = rs.rand(10, 5)
raw_Y = rs.rand(8, 5)
X = mt.tensor(raw_X)
Y = mt.tensor(raw_Y)
raw_sparse_x = sps.random(10,
5,
density=0.5,
format="csr",
random_state=rs)
raw_sparse_y = sps.random(8, 5, density=0.4, format="csr", random_state=rs)
X_sparse = mt.tensor(raw_sparse_x)
Y_sparse = mt.tensor(raw_sparse_y)
metric_func = lambda u, v: np.sqrt(((u - v)**2).sum())
_ = NearestNeighbors(algorithm="auto",
metric="precomputed",
metric_params={})
with pytest.raises(ValueError):
_ = NearestNeighbors(algorithm="unknown")
with pytest.raises(ValueError):
_ = NearestNeighbors(algorithm="kd_tree", metric=metric_func)
with pytest.raises(ValueError):
_ = NearestNeighbors(algorithm="auto", metric="unknown")
assert_warns(SyntaxWarning, NearestNeighbors, metric_params={"p": 1})
with pytest.raises(ValueError):
_ = NearestNeighbors(metric="wminkowski", p=0)
with pytest.raises(ValueError):
_ = NearestNeighbors(algorithm="auto", metric="minkowski", p=0)
nn = NearestNeighbors(algorithm="auto", metric="minkowski", p=1)
nn.fit(X)
assert nn.effective_metric_ == "manhattan"
nn = NearestNeighbors(algorithm="auto", metric="minkowski", p=2)
nn.fit(X)
assert nn.effective_metric_ == "euclidean"
nn = NearestNeighbors(algorithm="auto", metric="minkowski", p=np.inf)
nn.fit(X)
assert nn.effective_metric_ == "chebyshev"
nn2 = NearestNeighbors(algorithm="auto", metric="minkowski")
nn2.fit(nn)
assert nn2._fit_method == nn._fit_method
nn = NearestNeighbors(algorithm="auto", metric="minkowski")
ball_tree = SkBallTree(raw_X)
nn.fit(ball_tree)
assert nn._fit_method == "ball_tree"
nn = NearestNeighbors(algorithm="auto", metric="minkowski")
kd_tree = SkKDTree(raw_X)
nn.fit(kd_tree)
assert nn._fit_method == "kd_tree"
with pytest.raises(ValueError):
nn = NearestNeighbors()
nn.fit(np.random.rand(0, 10))
nn = NearestNeighbors(algorithm="ball_tree")
assert_warns(UserWarning, nn.fit, X_sparse)
nn = NearestNeighbors(metric="haversine")
with pytest.raises(ValueError):
nn.fit(X_sparse)
nn = NearestNeighbors(metric=metric_func, n_neighbors=1)
nn.fit(X)
assert nn._fit_method == "ball_tree"
nn = NearestNeighbors(metric="sqeuclidean", n_neighbors=1)
nn.fit(X)
assert nn._fit_method == "brute"
with pytest.raises(ValueError):
nn = NearestNeighbors(n_neighbors=-1)
nn.fit(X)
with pytest.raises(TypeError):
nn = NearestNeighbors(n_neighbors=1.3)
nn.fit(X)
nn = NearestNeighbors()
nn.fit(X)
with pytest.raises(ValueError):
nn.kneighbors(Y, n_neighbors=-1)
with pytest.raises(TypeError):
nn.kneighbors(Y, n_neighbors=1.3)
with pytest.raises(ValueError):
nn.kneighbors(Y, n_neighbors=11)
nn = NearestNeighbors(algorithm="ball_tree")
nn.fit(X)
with pytest.raises(ValueError):
nn.kneighbors(Y_sparse)