def _rebuild_covariance_matrix(self, covariance):
"""Rebuild the covariance matrix from its parameter values.
This method follows the steps:
* Rebuild a square matrix out of a triangular one.
* Add the missing half of the matrix by adding its transposed and
then removing the duplicated diagonal values.
* ensure the matrix is positive definite
Args:
covariance (list):
covariance values after unflattening model parameters.
Result:
list[list[float]]:
Symmetric positive semi-definite matrix.
"""
covariance = np.array(square_matrix(covariance))
covariance = (covariance + covariance.T -
(np.identity(covariance.shape[0]) * covariance))
if not check_matrix_symmetric_positive_definite(covariance):
covariance = make_positive_definite(covariance)
return covariance.tolist()
def test_make_positive_definite_iterate(mock_check):
"""Test find the nearest positive-definite matrix iterating."""
# Setup
mock_check.side_effect = [False, False, True]
# Run
matrix = np.array([[-1, -5], [-3, -7]])
result = make_positive_definite(matrix)
# Asserts
expected = np.array([[0.8, -0.4], [-0.4, 0.2]])
np.testing.assert_array_almost_equal(result, expected)
assert mock_check.call_count == 3
def test_make_positive_definite(mock_check):
"""Test find the nearest positive-definite matrix."""
# Setup
mock_check.return_value = True
# Run
matrix = np.array([[0, 1], [1, 0]])
result = make_positive_definite(matrix)
# Asserts
expected = np.array([[0.5, 0.5], [0.5, 0.5]])
np.testing.assert_equal(result, expected)
assert mock_check.call_count == 1
def _prepare_sampled_covariance(self, covariance):
"""Prepare a covariance matrix.
Args:
covariance (list):
covariance after unflattening model parameters.
Result:
list[list]:
symmetric Positive semi-definite matrix.
"""
covariance = np.array(square_matrix(covariance))
covariance = (covariance + covariance.T -
(np.identity(covariance.shape[0]) * covariance))
if not check_matrix_symmetric_positive_definite(covariance):
covariance = make_positive_definite(covariance)
return covariance.tolist()