def __init__(self,
power=0.,
validate_args=False,
parameters=None,
name='power_transform'):
"""Instantiates the `PowerTransform` bijector.
Args:
power: Python `float` scalar indicating the transform power, i.e.,
`Y = g(X) = (1 + X * c)**(1 / c)` where `c` is the `power`.
validate_args: Python `bool` indicating whether arguments should be
checked for correctness.
parameters: Locals dict captured by subclass constructor, to be used for
copy/slice re-instantiation operators.
name: Python `str` name given to ops managed by this object.
Raises:
ValueError: if `power < 0` or is not known statically.
"""
parameters = dict(locals()) if parameters is None else parameters
with tf.name_scope(name) as name:
power = tf.get_static_value(ps.constant(power, name='power'))
if power is None or power < 0:
raise ValueError('`power` must be a non-negative TF constant.')
self._power = power
super(PowerTransform, self).__init__(forward_min_event_ndims=0,
validate_args=validate_args,
parameters=parameters,
name=name)
def _event_shape_tensor(self, index_points=None):
index_points = self._get_index_points(index_points)
if self._is_univariate_marginal(index_points):
return ps.constant([], dtype=tf.int32)
else:
# The examples index is one position to the left of the feature dims.
examples_index = -(self.kernel.feature_ndims + 1)
return ps.shape(index_points)[examples_index:examples_index + 1]
def _event_shape_tensor(self, loc=None):
del loc
return ps.constant([], dtype=tf.int32)
def test_convert_dimension_to_tensor(self):
v = ps.constant(tf1.Dimension(1))
self.assertEqual(1, v)