def _parameter_properties(cls, dtype, num_classes=None):
return dict(
index_points=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims + 1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED,
),
observations=parameter_properties.ParameterProperties(
event_ndims=2,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED),
observation_index_points=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims + 1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED,
),
observations_is_missing=parameter_properties.ParameterProperties(
event_ndims=2,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED,
),
kernel=parameter_properties.BatchedComponentProperties(),
observation_noise_variance=parameter_properties.
ParameterProperties(
event_ndims=0,
shape_fn=lambda sample_shape: sample_shape[:-1],
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype
)))),
predictive_noise_variance=parameter_properties.ParameterProperties(
event_ndims=0,
shape_fn=lambda sample_shape: sample_shape[:-1],
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype
)))),
_observation_scale=parameter_properties.BatchedComponentProperties(
))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
loc=parameter_properties.ParameterProperties(event_ndims=1),
precision_factor=parameter_properties.BatchedComponentProperties(),
precision=parameter_properties.BatchedComponentProperties(),
nonzeros=parameter_properties.BatchedComponentProperties(
event_ndims=1))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
mixture_distribution=(
parameter_properties.BatchedComponentProperties()),
components_distribution=(
parameter_properties.BatchedComponentProperties(
event_ndims=1)))
def _parameter_properties(cls, dtype):
return dict(
chain=parameter_properties.BatchedComponentProperties(),
transition_bijector=parameter_properties.BatchedComponentProperties(
# The transition bijector contributes no batch shape
# beyond that from the chain itself.
event_ndims=None))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
bijector=parameter_properties.BatchedComponentProperties(
event_ndims=lambda td: tf.nest.map_structure( # pylint: disable=g-long-lambda
tensorshape_util.rank, td.distribution.event_shape),
event_ndims_tensor=lambda td: tf.nest.map_structure( # pylint: disable=g-long-lambda
ps.rank_from_shape, td.distribution.event_shape_tensor())))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
df=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype)))),
loc=parameter_properties.ParameterProperties(event_ndims=2),
scale_row=parameter_properties.BatchedComponentProperties(),
scale_column=parameter_properties.BatchedComponentProperties())
def _parameter_properties(cls, dtype, num_classes=None):
# pylint: disable=g-long-lambda
return dict(df=parameter_properties.ParameterProperties(
shape_fn=lambda sample_shape: sample_shape[:-2],
default_constraining_bijector_fn=parameter_properties.
BIJECTOR_NOT_IMPLEMENTED),
scale=parameter_properties.BatchedComponentProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(
# TODO(davmre): replace with `self.reinterpreted_batch_ndims` once
# support for `reinterpreted_batch_ndims=None` has been removed.
event_ndims=lambda self: self._get_reinterpreted_batch_ndims()), # pylint: disable=protected-access
reinterpreted_batch_ndims=(
parameter_properties.ShapeParameterProperties()))
def _parameter_properties(cls, dtype):
from tensorflow_probability.python.bijectors import softplus # pylint:disable=g-import-not-at-top
return dict(
kernel=parameter_properties.BatchedComponentProperties(),
scale_diag=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims,
default_constraining_bijector_fn=(
lambda: softplus.Softplus(low=dtype_util.eps(dtype)))))
def _parameter_properties(cls, dtype, num_classes=None):
from tensorflow_probability.python.bijectors import softplus as softplus_bijector # pylint:disable=g-import-not-at-top
return dict(
amplitudes=parameter_properties.ParameterProperties(
event_ndims=1,
default_constraining_bijector_fn=(
softplus_bijector.Softplus(low=dtype_util.eps(dtype)))),
kernel=parameter_properties.BatchedComponentProperties(event_ndims=1))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
low=parameter_properties.ParameterProperties(),
# TODO(b/169874884): Support decoupled parameterization.
high=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=parameter_properties.
BIJECTOR_NOT_IMPLEMENTED, ))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
df=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype)))),
index_points=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims + 1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED),
kernel=parameter_properties.BatchedComponentProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
df=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus( # pylint:disable=g-long-lambda
low=dtype_util.as_numpy_dtype(dtype)(2.)))),
schur_complement=parameter_properties.BatchedComponentProperties(),
fixed_inputs_observations=parameter_properties.ParameterProperties(
event_ndims=1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED))
def _parameter_properties(cls, dtype):
from tensorflow_probability.python.bijectors import softplus # pylint:disable=g-import-not-at-top
return dict(
base_kernel=parameter_properties.BatchedComponentProperties(),
fixed_inputs=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.base_kernel.feature_ndims + 1),
diag_shift=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus.Softplus(low=dtype_util.eps(dtype)))),
_precomputed_divisor_matrix_cholesky=(
parameter_properties.ParameterProperties(event_ndims=2)))
def _parameter_properties(cls, dtype, num_classes=None):
# Subclasses must implement their own `_parameter_properties`. If they
# don't, call the base Distribution version to raise a NotImplementedError.
if cls is MultivariateStudentTLinearOperator:
return dict(
df=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(
dtype)))),
loc=parameter_properties.ParameterProperties(event_ndims=1),
scale=parameter_properties.BatchedComponentProperties())
return distribution.Distribution._parameter_properties(dtype=dtype) # pylint: disable=protected-access
def _parameter_properties(cls, dtype):
def get_parameter_event_ndims(self, x_event_ndims):
return nest.map_structure_up_to(
self.
bijectors, # Recurse up to the BijectorWithMetadata tuples.
lambda bm: bm.x_event_ndims,
self._get_bijectors_with_metadata( # pylint: disable=protected-access
event_ndims=x_event_ndims,
pack_as_original_structure=True))
return dict(bijectors=parameter_properties.BatchedComponentProperties(
event_ndims=get_parameter_event_ndims))
def _parameter_properties(cls, dtype, num_classes=None):
from tensorflow_probability.python.bijectors import softplus as softplus_bijector # pylint:disable=g-import-not-at-top
return dict(index_points=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims + 1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED),
kernel=parameter_properties.BatchedComponentProperties(),
observation_noise_variance=parameter_properties.
ParameterProperties(
event_ndims=0,
shape_fn=lambda sample_shape: sample_shape[:-1],
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.
eps(dtype)))))
def _parameter_properties(cls, dtype, num_classes=None):
# pylint: disable=g-long-lambda
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
shift=parameter_properties.ParameterProperties(),
scale=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype
)))),
tailweight=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype
)))))
def _parameter_properties(cls, dtype, num_classes=None):
from tensorflow_probability.python.bijectors import ascending # pylint:disable=g-import-not-at-top
from tensorflow_probability.python.bijectors import softplus # pylint:disable=g-import-not-at-top
return dict(
kernels=parameter_properties.BatchedComponentProperties(
event_ndims=lambda self: [0 for _ in self.kernels]),
locs=parameter_properties.ParameterProperties(
event_ndims=1,
default_constraining_bijector_fn=lambda: ascending.Ascending(
)), # pylint:disable=unnecessary-lambda
slopes=parameter_properties.ParameterProperties(
event_ndims=1,
default_constraining_bijector_fn=(
lambda: softplus.Softplus(low=dtype_util.eps(dtype)))))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
df=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus( # pylint: disable=g-long-lambda
low=dtype_util.as_numpy_dtype(dtype)(2.)))),
index_points=parameter_properties.ParameterProperties(
event_ndims=lambda self: self.kernel.feature_ndims + 1,
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED),
kernel=parameter_properties.BatchedComponentProperties(),
observation_noise_variance=parameter_properties.ParameterProperties(
default_constraining_bijector_fn=(
lambda: softplus_bijector.Softplus(low=dtype_util.eps(dtype))),
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distributions=parameter_properties.BatchedComponentProperties(
event_ndims=lambda self: [0 for _ in self.distributions]),
axis=parameter_properties.ShapeParameterProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
sample_shape=parameter_properties.ShapeParameterProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(
event_ndims=lambda self: self.reinterpreted_batch_ndims))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
loc=parameter_properties.ParameterProperties(event_ndims=1),
scale=parameter_properties.BatchedComponentProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
bijector=parameter_properties.BatchedComponentProperties(
event_ndims=_count_bijector_batch_ndims_used_by_event))
def _parameter_properties(cls, dtype):
return dict(bijector=parameter_properties.BatchedComponentProperties(
event_ndims=(
lambda self, x_event_ndims: self.bijector.inverse_event_ndims( # pylint: disable=g-long-lambda
x_event_ndims))))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
base_kernel=parameter_properties.BatchedComponentProperties(),
task_kernel_matrix_linop=(
parameter_properties.BatchedComponentProperties()))
def _parameter_properties(cls, dtype):
return dict(
bijectors=parameter_properties.BatchedComponentProperties(
event_ndims=lambda self: [None for _ in self.bijectors]))
def _parameter_properties(cls, dtype, num_classes=None):
return dict(kernel=parameter_properties.BatchedComponentProperties())
def _parameter_properties(cls, dtype, num_classes=None):
return dict(
distribution=parameter_properties.BatchedComponentProperties(),
validity_mask=parameter_properties.ParameterProperties(
shape_fn=parameter_properties.SHAPE_FN_NOT_IMPLEMENTED))
