def test_std_with_ddof(self):
x = np.random.uniform(0, 10, (20, 100))
for axis in [None, 0, 1]:
np.testing.assert_almost_equal(
np.std(x, axis=axis, ddof=10),
std(csr_matrix(x), axis=axis, ddof=10),
)
def test_std_with_ddof(self):
x = np.random.uniform(0, 10, (20, 100))
for axis in [None, 0, 1]:
np.testing.assert_almost_equal(
np.std(x, axis=axis, ddof=10),
std(csr_matrix(x), axis=axis, ddof=10),
)
def _standardize(X, tol=1e-8):
""" Standardize matrix X avoiding NaNs and small values. """
s = ut.std(X, axis=0)
m = np.zeros(*s.shape)
valid = s > tol
m[np.logical_not(valid)] = 0
m[valid] = 1.0 / s[valid]
return (X - X.mean(axis=0)) * m
def test_std(self):
for data in self.data:
for axis in chain((None,), range(len(data.shape))):
# Can't use array_equal here due to differences on 1e-16 level
np.testing.assert_array_almost_equal(
std(csr_matrix(data), axis=axis),
np.std(data, axis=axis)
)
def test_std(self):
for data in self.data:
for axis in chain((None,), range(len(data.shape))):
# Can't use array_equal here due to differences on 1e-16 level
np.testing.assert_array_almost_equal(
std(csr_matrix(data), axis=axis),
np.std(data, axis=axis)
)
