def test_default_seed_is_zero(self):
n_models = 2
n_features = 3
cepstral1 = VarianceRegressor(n_models=n_models, n_features=n_features)
cepstral2 = VarianceRegressor(n_models=n_models,
n_features=n_features,
rng=0)
np.testing.assert_allclose(cepstral1.prediction_error.weights_,
cepstral2.prediction_error.weights_)
def test_rng_can_be_seed(self):
n_models = 2
n_features = 3
seed = 43
cepstral1 = VarianceRegressor(n_models=n_models,
n_features=n_features,
rng=seed)
cepstral2 = VarianceRegressor(
n_models=n_models,
n_features=n_features,
rng=np.random.default_rng(seed),
)
np.testing.assert_allclose(cepstral1.prediction_error.weights_,
cepstral2.prediction_error.weights_)
def setUp(self):
self.n_models = 3
self.n_features = 4
self.order = 3
self.rng = np.random.default_rng(1)
self.weights = self.rng.normal(size=(self.n_models, self.n_features))
self.variance = VarianceRegressor(self.weights)
def test_weights_initialized_using_rng_normal(self):
n_models = 2
n_features = 3
rng = mock.Mock()
weights = np.reshape(np.arange(n_models * n_features),
(n_models, n_features))
rng.normal = mock.Mock(return_value=weights)
cepstral = VarianceRegressor(n_models=n_models,
n_features=n_features,
rng=rng)
np.testing.assert_allclose(cepstral.prediction_error.weights_, weights)
def test_raises_if_n_features_does_not_match_weights_size(self):
with self.assertRaises(ValueError):
VarianceRegressor(n_features=2, weights=np.zeros((3, 4)))