def test_intrinsic_dim_mle_levina_low_memory(self):
""" Same as above, but invoking the speed-memory trade-off. """
_, _, vector = load_dexter()
ID_MLE_REF = 74.472
id_mle = intrinsic_dimension(vector, 6, 12, 'levina',
'vector', None, mem_threshold=0)
return np.testing.assert_almost_equal(id_mle, ID_MLE_REF, decimal=3)
def test_intrinsic_dim_mle_levina(self):
"""Test against value calc. by matlab reference implementation."""
_, _, vector = load_dexter()
ID_MLE_REF = 74.472
id_mle = intrinsic_dimension(vector, k1=6, k2=12,
estimator='levina', metric='vector', trafo=None)
return np.testing.assert_almost_equal(id_mle, ID_MLE_REF, decimal=3)
def _calc_intrinsic_dim(self):
"""Calculate intrinsic dimension estimate."""
self.intrinsic_dim = intrinsic_dimension(X=self.vectors)
return self
def test_incorrect_metric_other(self):
""" Test handling of unsupported metric parameters."""
with self.assertRaises(ValueError):
intrinsic_dimension(self.vector, metric=None)
def test_incorrect_metric_sim(self):
""" Test handling of unsupported metric parameters."""
with self.assertRaises(NotImplementedError):
intrinsic_dimension(self.vector, metric='similarity')
def test_incorrect_trafo_params(self):
""" Test handling of incorrect transformation parameters."""
with self.assertRaises(ValueError):
intrinsic_dimension(self.vector, trafo=0)
def test_incorrect_k2_params(self):
""" Test handling of incorrect neighborhood parameters."""
n = self.vector.shape[0]
with self.assertRaises(ValueError):
intrinsic_dimension(self.vector, k2=n)
def test_incorrect_est_params(self):
""" Test handling of incorrect estimator. """
with self.assertRaises(ValueError):
intrinsic_dimension(self.vector,
estimator='the_single_truly_best_id_estimator')