# Copyright (c) 2017-2019 Uber Technologies, Inc.
# SPDX-License-Identifier: Apache-2.0
import torch
from torch.distributions import constraints
import pyro
import pyro.distributions as dist
from pyro.distributions.util import is_identically_one, is_validation_enabled
from .reparam import Reparam
[docs]class LocScaleReparam(Reparam):
"""
Generic decentering reparameterizer [1] for latent variables parameterized
by ``loc`` and ``scale`` (and possibly additional ``shape_params``).
This reparameterization works only for latent variables, not likelihoods.
[1] Maria I. Gorinova, Dave Moore, Matthew D. Hoffman (2019)
"Automatic Reparameterisation of Probabilistic Programs"
https://arxiv.org/pdf/1906.03028.pdf
:param float centered: optional centered parameter. If None (default) learn
a per-site per-element centering parameter in ``[0,1]``. If 0, fully
decenter the distribution; if 1, preserve the centered distribution
unchanged.
:param shape_params: Optional list of additional parameter names to copy
unchanged from the centered to decentered distribution. If absent,
all params in a distributions ``.arg_constraints`` will be copied.
:type shape_params: tuple or list
"""
def __init__(self, centered=None, shape_params=None):
assert centered is None or isinstance(centered, (float, torch.Tensor))
if shape_params is not None:
assert isinstance(shape_params, (tuple, list))
assert all(isinstance(name, str) for name in shape_params)
if is_validation_enabled():
if isinstance(centered, float):
assert 0 <= centered and centered <= 1
elif isinstance(centered, torch.Tensor):
assert (0 <= centered).all()
assert (centered <= 1).all()
else:
assert centered is None
self.centered = centered
self.shape_params = shape_params
[docs] def apply(self, msg):
name = msg["name"]
fn = msg["fn"]
value = msg["value"]
is_observed = msg["is_observed"]
centered = self.centered
if is_identically_one(centered):
return msg
event_shape = fn.event_shape
fn, event_dim = self._unwrap(fn)
# Apply a partial decentering transform.
if self.shape_params is None:
self.shape_params = tuple(
k for k in fn.arg_constraints if k not in ("loc", "scale")
)
params = {key: getattr(fn, key) for key in self.shape_params}
if centered is None:
centered = pyro.param(
"{}_centered".format(name),
lambda: fn.loc.new_full(event_shape, 0.5),
constraint=constraints.unit_interval,
)
params["loc"] = fn.loc * centered
params["scale"] = fn.scale**centered
decentered_fn = type(fn)(**params)
# Differentiably invert transform.
decentered_value = None
if value is not None:
delta = (value - fn.loc) * fn.scale.pow(centered - 1)
decentered_value = delta + centered * fn.loc
# Draw decentered noise.
decentered_value = pyro.sample(
f"{name}_decentered",
self._wrap(decentered_fn, event_dim),
obs=decentered_value,
infer={"is_observed": is_observed},
)
# Differentiably transform.
if value is None:
delta = decentered_value - centered * fn.loc
value = fn.loc + fn.scale.pow(1 - centered) * delta
# Simulate a pyro.deterministic() site.
new_fn = dist.Delta(value, event_dim=event_dim).mask(False)
return {"fn": new_fn, "value": value, "is_observed": True}