def test_smoke_ccc(self):
"""Smoke test for ccc MI."""
for cop in [True, False]:
for cat in [None, categories]:
# tensor-based
c = GCMIEstimator(mi_type='ccc', tensor=True, copnorm=cop)
mi_t = c.estimate(x, y_c, z=z_c, categories=cat)
# vector-based
c = GCMIEstimator(mi_type='ccc', tensor=False, copnorm=cop)
mi_v = c.estimate(x, y_c, z=z_c, categories=cat)
np.testing.assert_array_almost_equal(mi_t, mi_v)
def test_support_dim(self):
"""Test the support for different dimensions."""
y = np.random.rand(100)
c = GCMIEstimator(mi_type='cc')
# 1d
x = np.random.rand(100)
c = GCMIEstimator(mi_type='cc')
mi = c.estimate(x, y)
assert mi.shape == (1, 1)
# 2d
x = np.random.rand(1, 100)
mi = c.estimate(x, y)
assert mi.shape == (1, 1)
# Nd
x = np.random.rand(4, 5, 6, 1, 100)
mi = c.estimate(x, y)
assert mi.shape == (1, 4, 5, 6)
def test_functional_cc(self):
"""Functional test for cc MI."""
# build the testing data
x = np.random.rand(n_times, n_mv, n_samples)
y = np.random.normal(size=(1, 1, n_samples,))
x[sl_effect, :, :] += y
y = np.tile(y, (n_times, 1, 1))
# functional test
for cop in [True, False]:
for imp in [True, False]:
c = GCMIEstimator(mi_type='cc', tensor=imp, copnorm=cop)
mi = c.estimate(x, y)
self._compare_effects(effect, mi)
def test_functional_cd(self):
"""Functional test for cd MI."""
# build the testing data
x = np.random.rand(n_times, n_mv, n_samples)
x[sl_effect, :, 0:50] += 10.
x[sl_effect, :, 50::] -= 10.
y = np.array([0] * 50 + [1] * 50)
# functional test
for cop in [True, False]:
for imp in [True, False]:
c = GCMIEstimator(mi_type='cd', tensor=imp, copnorm=cop)
mi = c.estimate(x, y)
self._compare_effects(effect, mi)
def test_functional_ccd(self):
"""Functional test for ccd MI."""
# build the testing data
x = np.random.rand(n_times, n_mv, n_samples)
y_pos = np.random.normal(size=(1, 1, 50))
y_neg = np.random.normal(size=(1, 1, 50))
x[sl_effect, :, 0:50] += y_pos
x[sl_effect, :, 50::] -= y_neg
y = np.tile(np.concatenate((y_pos, y_neg), axis=2), (n_times, 1, 1))
z = np.array([0] * 50 + [1] * 50)
# functional test
for cop in [True, False]:
for imp in [True, False]:
c = GCMIEstimator(mi_type='ccd', tensor=imp, copnorm=cop)
mi = c.estimate(x, y, z=z)
self._compare_effects(effect, mi)