def test_using_previously_not_fitted_scaler(self):
scaler = StandardScaler()
original_shape = (25, 10, 7)
X = np.random.rand(*original_shape)
with self.assertRaises(NotFittedError):
dataset.normalise(X, scaler, fit_scaler=False)
def test_not_normalising_all_columns(self):
scaler = StandardScaler()
original_shape = (25, 10, 7)
columns_to_normalise_bool_index = np.array(
[False for _ in range(original_shape[-1])])
X = np.random.rand(*original_shape)
with self.assertRaises(ValueError):
dataset.normalise(X, scaler, True, columns_to_normalise_bool_index)
def test_using_previously_fitted_scaler(self):
scaler = StandardScaler()
original_shape = (25, 10, 7)
X = np.random.rand(*original_shape)
X = dataset.normalise(X, scaler, fit_scaler=True)
scale_ = scaler.scale_
mean_ = scaler.mean_
n_samples_seen_ = scaler.n_samples_seen_
X2 = np.random.rand(50, 4, 7)
X2 = dataset.normalise(X, scaler, fit_scaler=False)
self.assertTrue(np.isclose(scaler.scale_, scale_).all())
self.assertTrue(np.isclose(scaler.mean_, mean_).all())
self.assertEqual(scaler.n_samples_seen_, n_samples_seen_)
def test_fitting_scaler(self):
scaler = StandardScaler()
original_shape = (25, 10, 7)
X = np.random.rand(*original_shape)
X = dataset.normalise(X, scaler, fit_scaler=True)
self.assertTrue(self.is_shape_correct(original_shape, X))
self.assertTrue(self.are_mean_and_variance_correct(X))
def test_not_normalising_some_columns(self):
scaler = StandardScaler()
original_shape = (25, 10, 7)
columns_to_normalise_bool_index = np.array(
[True, True, False, True, False, True, True])
original_X = np.random.rand(*original_shape)
X = dataset.normalise(original_X, scaler, True,
columns_to_normalise_bool_index)
self.assertTrue(
np.isclose(original_X[:, :, columns_to_normalise_bool_index],
X[:, :, columns_to_normalise_bool_index]).all())
self.assertTrue(
self.are_mean_and_variance_correct(
X[:, :, columns_to_normalise_bool_index]))