def transform(self, X):
X = resize_x(X)
X_msc = np.zeros_like(X)
for i in range(X.shape[0]):
fit = np.polyfit(self.reference, X[i, :], 1, full=True)
X_msc[i, :] = np.divide((X[i, :] - fit[0][1]), fit[0][0])
return X_msc
def transform(self, X: np.array) -> np.array:
"""
Remove the mean row wise.
:param X: numpy array.
:return: numpy array.
"""
X_val = deepcopy(X)
X_val = resize_x(X_val)
mean = np.tile(np.mean(X_val, axis=1), (X_val.shape[1], 1)).T
X_mean_centering = X_val - mean
return X_mean_centering
def transform(self, X: np.array) -> np.array:
"""
Remove the mean and reduce by the standard deviation row wise.
:param X: numpy array.
:return: numpy array.
"""
X_val = deepcopy(X)
X_val = resize_x(X_val)
mean = np.tile(np.mean(X_val, axis=1), (X_val.shape[1], 1)).T
std = np.tile(np.std(X_val, axis=1), (X_val.shape[1], 1)).T
X_snv = (X_val - mean) / std
return X_snv
def transform(self, X):
X_val = resize_x(X)
if self.norm == "l1":
norm_matrix = np.tile(
np.linalg.norm(X_val, ord=1, axis=1, keepdims=True), (1, X_val.shape[1])
)
elif self.norm == "l2":
norm_matrix = np.tile(
np.linalg.norm(X_val, ord=2, axis=1, keepdims=True), (1, X_val.shape[1])
)
elif self.norm == "inf":
norm_matrix = np.tile(
np.linalg.norm(X_val, ord=np.inf, axis=1, keepdims=True),
(1, X_val.shape[1]),
)
X_norm = np.divide(X_val, norm_matrix)
return X_norm
def fit(self, X, y=None):
"""
Set the reference spectrum
"""
X_val = resize_x(X)
if X_val.shape[0] == 1 and y is None:
raise ValueError(
"A reference spectrum y must be given for X with only one row."
)
if y is None:
self.reference = np.mean(X, axis=0)
else:
if y.shape != (1, X_val.shape[1]):
raise ValueError(
f"The reference must be of shape {(1, X_val.shape[1])} but is ({y.shape})"
)
else:
self.reference = y
return self
def test_resize_x_1d(self):
x = np.array([0, 1, 2, 3])
resized = resize_x(x)
np.testing.assert_almost_equal(resized, np.array([[0, 1, 2, 3]]))
def test_resize_x_2d(self):
x = np.array([[0, 1], [2, 3]])
resized = resize_x(x)
np.testing.assert_almost_equal(x, resized)
def test_resize_x_fail(self):
with pytest.raises(ArrayDimensionError):
resize_x(np.array([[[0], [1], [2]], [[0], [1], [2]]]))