in tensorflow_probability/python/experimental/mcmc/initialization.py [0:0]
def init_near_unconstrained_zero(
model=None, constraining_bijector=None, event_shapes=None,
event_shape_tensors=None, batch_shapes=None, batch_shape_tensors=None,
dtypes=None, shard_axis_names=None):
"""Returns an initialization Distribution for starting a Markov chain.
This initialization scheme follows Stan: we sample every latent
independently, uniformly from -2 to 2 in its unconstrained space,
and then transform into constrained space to construct an initial
state that can be passed to `sample_chain` or other MCMC drivers.
The argument signature is arranged to let the user pass either a
`JointDistribution` describing their model, if it's in that form, or
the essential information necessary for the sampling, namely a
bijector (from unconstrained to constrained space) and the desired
shape and dtype of each sample (specified in constrained space).
Note: As currently implemented, this function has the limitation
that the batch shape of the supplied model is ignored, but that
could probably be generalized if needed.
Args:
model: A `Distribution` (typically a `JointDistribution`) giving the
model to be initialized. If supplied, it is queried for
its default event space bijector, its event shape, and its dtype.
If not supplied, those three elements must be supplied instead.
constraining_bijector: A (typically multipart) `Bijector` giving
the mapping from unconstrained to constrained space. If
supplied together with a `model`, acts as an override. A nested
structure of `Bijector`s is accepted, and interpreted as
applying in parallel to a corresponding structure of state parts
(see `JointMap` for details).
event_shapes: A structure of shapes giving the (unconstrained)
event space shape of the desired samples. Must be an acceptable
input to `constraining_bijector.inverse_event_shape`. If
supplied together with `model`, acts as an override.
event_shape_tensors: A structure of tensors giving the (unconstrained)
event space shape of the desired samples. Must be an acceptable
input to `constraining_bijector.inverse_event_shape_tensor`. If
supplied together with `model`, acts as an override. Required if any of
`event_shapes` are not fully-defined.
batch_shapes: A structure of shapes giving the batch shape of the desired
samples. If supplied together with `model`, acts as an override. If
unspecified, we assume scalar batch `[]`.
batch_shape_tensors: A structure of tensors giving the batch shape of the
desired samples. If supplied together with `model`, acts as an override.
Required if any of `batch_shapes` are not fully-defined.
dtypes: A structure of dtypes giving the (unconstrained) dtypes of
the desired samples. Must be an acceptable input to
`constraining_bijector.inverse_dtype`. If supplied together
with `model`, acts as an override.
shard_axis_names: A structure of `str`s indicating the named axes by which
the distribution event is sharded. See
`tfp.experimental.distribute.Sharded` for more context.
Returns:
init_dist: A `Distribution` representing the initialization
distribution, in constrained space. Samples from this
`Distribution` are valid initial states for a Markov chain
targeting the model.
#### Example
Initialize 100 chains from the unconstrained -2, 2 distribution
for a model expressed as a `JointDistributionCoroutine`:
```python
@tfp.distributions.JointDistributionCoroutine
def model():
...
init_dist = tfp.experimental.mcmc.init_near_unconstrained_zero(model)
states = tfp.mcmc.sample_chain(
current_state=init_dist.sample(100, seed=[4, 8]),
...)
```
"""
# Canonicalize arguments into the parts we need, namely
# the constraining_bijector, the event_shapes, and the dtypes.
if model is not None:
# Got a Distribution model; treat other arguments as overrides if
# present.
if constraining_bijector is None:
# pylint: disable=protected-access
constraining_bijector = model.experimental_default_event_space_bijector()
if event_shapes is None:
event_shapes = model.event_shape
if event_shape_tensors is None:
event_shape_tensors = model.event_shape_tensor()
if dtypes is None:
dtypes = model.dtype
if batch_shapes is None:
batch_shapes = nest_util.broadcast_structure(dtypes, model.batch_shape)
if batch_shape_tensors is None:
batch_shape_tensors = nest_util.broadcast_structure(
dtypes, model.batch_shape_tensor())
if shard_axis_names is None:
shard_axis_names = model.experimental_shard_axis_names
else:
if constraining_bijector is None or event_shapes is None or dtypes is None:
msg = ('Must pass either a Distribution (typically a JointDistribution), '
'or a bijector, a structure of event shapes, and a '
'structure of dtypes')
raise ValueError(msg)
event_shapes_fully_defined = all(tensorshape_util.is_fully_defined(s)
for s in tf.nest.flatten(event_shapes))
if not event_shapes_fully_defined and event_shape_tensors is None:
raise ValueError('Must specify `event_shape_tensors` when `event_shapes` '
f'are not fully-defined: {event_shapes}')
if batch_shapes is None:
batch_shapes = tf.TensorShape([])
batch_shapes = nest_util.broadcast_structure(dtypes, batch_shapes)
batch_shapes_fully_defined = all(tensorshape_util.is_fully_defined(s)
for s in tf.nest.flatten(batch_shapes))
if batch_shape_tensors is None:
if not batch_shapes_fully_defined:
raise ValueError(
'Must specify `batch_shape_tensors` when `batch_shapes` are not '
f'fully-defined: {batch_shapes}')
batch_shape_tensors = tf.nest.map_structure(
tf.convert_to_tensor, batch_shapes)
# Interpret a structure of Bijectors as the joint multipart bijector.
if not isinstance(constraining_bijector, tfb.Bijector):
constraining_bijector = tfb.JointMap(constraining_bijector)
# Actually initialize
def one_term(event_shape, event_shape_tensor, batch_shape, batch_shape_tensor,
dtype, shard_axes=None):
if not tensorshape_util.is_fully_defined(event_shape):
event_shape = event_shape_tensor
result = tfd.Sample(
tfd.Uniform(low=tf.constant(-2., dtype=dtype),
high=tf.constant(2., dtype=dtype)),
sample_shape=event_shape)
if shard_axes:
result = distribute.Sharded(result, shard_axes)
if not tensorshape_util.is_fully_defined(batch_shape):
batch_shape = batch_shape_tensor
needs_bcast = True
else: # Only batch broadcast when batch ndims > 0.
needs_bcast = bool(tensorshape_util.as_list(batch_shape))
if needs_bcast:
result = tfd.BatchBroadcast(result, batch_shape)
return result
inv_shapes = constraining_bijector.inverse_event_shape(event_shapes)
if event_shape_tensors is not None:
inv_shape_tensors = constraining_bijector.inverse_event_shape_tensor(
event_shape_tensors)
else:
inv_shape_tensors = tf.nest.map_structure(lambda _: None, inv_shapes)
inv_dtypes = constraining_bijector.inverse_dtype(dtypes)
if shard_axis_names is None:
shard_axis_names = tf.nest.map_structure(lambda _: None, batch_shapes)
terms = nest.map_structure_up_to(inv_shapes, one_term, inv_shapes,
inv_shape_tensors, batch_shapes,
batch_shape_tensors,
inv_dtypes, shard_axis_names)
if shard_axis_names and any(shard_axes for shard_axes in nest.flatten_up_to(
batch_shapes, shard_axis_names)):
dist = distribute.JointDistributionSequential(tf.nest.flatten(terms))
else:
dist = tfd.JointDistributionSequential(tf.nest.flatten(terms))
unconstrained = tfb.pack_sequence_as(inv_shapes)(dist)
return tfd.TransformedDistribution(
unconstrained, bijector=constraining_bijector)