def test_repeat_array():
n_elements = 10
x = np.zeros(n_elements)
# repeating row-wise
r1 = utils.repeat_array(x, length=1, axis=0)
r2 = utils.repeat_array(x, length=2, axis=0)
assert r1.shape == (1, n_elements)
assert r2.shape == (2, n_elements)
# repeating column-wise
c1 = utils.repeat_array(x, length=1, axis=1)
c2 = utils.repeat_array(x, length=2, axis=1)
assert c1.shape == (n_elements, 1)
assert c2.shape == (n_elements, 2)
def transform(self, x: ArrayLike) -> np.ndarray:
"""Scale covariates according to the feature range.
Args:
x: array-like of shape (n_samples, n_features)
Input data that will be transformed.
Returns:
ndarray with transformed data.
"""
x = np.array(x)
xarr = repeat_array(x, len(self.threshold_indices_), axis=-1)
tarr = repeat_array(self.threshold_indices_.transpose(),
len(x),
axis=0)
thresh = (xarr > tarr).reshape(x.shape[0], -1)
return thresh.astype(np.uint8)
def transform(self, x: ArrayLike) -> np.ndarray:
"""Scale covariates according to the feature range.
Args:
x: array-like of shape (n_samples, n_features)
Input data that will be transformed.
Returns:
ndarray with transformed data.
"""
x = np.array(x)
xarr = repeat_array(x, self.n_hinges_ - 1, axis=-1)
lharr = repeat_array(self.hinge_indices_[:-1].transpose(),
len(x),
axis=0)
rharr = repeat_array(self.hinge_indices_[1:].transpose(),
len(x),
axis=0)
lh = left_hinge(xarr, lharr, self.maxs_)
rh = right_hinge(xarr, self.mins_, rharr)
return np.concatenate((lh, rh), axis=2).reshape(x.shape[0], -1)
def right_hinge(x: ArrayLike, mn: float, mx: float) -> np.ndarray:
"""Computes hinge transformation values.
Args:
x: Array-like of covariate values
mn: Minimum covariate value to fit hinges to
mx: Maximum covariate value to fit hinges to
Returns:
Array of hinge features
"""
mn_broadcast = repeat_array(mn, mx.shape[-1], axis=1)
return np.minimum(1, np.maximum(0,
(x - mn_broadcast) / (mx - mn_broadcast)))
def left_hinge(x: ArrayLike, mn: float, mx: float) -> np.ndarray:
"""Computes hinge transformation values.
Args:
x: Array-like of covariate values
mn: Minimum covariate value to fit hinges to
mx: Maximum covariate value to fit hinges to
Returns:
Array of hinge features
"""
return np.minimum(
1,
np.maximum(0,
(x - mn) / (repeat_array(mx, mn.shape[-1], axis=1) - mn)))