# Copyright 2016 The TensorFlow Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== """Bijector base.""" import abc import collections import contextlib import re import numpy as np from tensorflow.python.framework import dtypes from tensorflow.python.framework import ops from tensorflow.python.framework import tensor_shape from tensorflow.python.framework import tensor_util from tensorflow.python.ops import array_ops from tensorflow.python.ops import check_ops from tensorflow.python.ops import math_ops from tensorflow.python.ops.distributions import util as distribution_util from tensorflow.python.util import object_identity __all__ = [ "Bijector", ] class _Mapping(collections.namedtuple( "_Mapping", ["x", "y", "ildj_map", "kwargs"])): """Helper class to make it easier to manage caching in `Bijector`.""" def __new__(cls, x=None, y=None, ildj_map=None, kwargs=None): """Custom __new__ so namedtuple items have defaults. Args: x: `Tensor`. Forward. y: `Tensor`. Inverse. ildj_map: `Dictionary`. This is a mapping from event_ndims to a `Tensor` representing the inverse log det jacobian. kwargs: Python dictionary. Extra args supplied to forward/inverse/etc functions. Returns: mapping: New instance of _Mapping. """ return super(_Mapping, cls).__new__(cls, x, y, ildj_map, kwargs) @property def x_key(self): """Returns key used for caching Y=g(X).""" return ((object_identity.Reference(self.x),) + self._deep_tuple(tuple(sorted(self.kwargs.items())))) @property def y_key(self): """Returns key used for caching X=g^{-1}(Y).""" return ((object_identity.Reference(self.y),) + self._deep_tuple(tuple(sorted(self.kwargs.items())))) def merge(self, x=None, y=None, ildj_map=None, kwargs=None, mapping=None): """Returns new _Mapping with args merged with self. Args: x: `Tensor`. Forward. y: `Tensor`. Inverse. ildj_map: `Dictionary`. This is a mapping from event_ndims to a `Tensor` representing the inverse log det jacobian. kwargs: Python dictionary. Extra args supplied to forward/inverse/etc functions. mapping: Instance of _Mapping to merge. Can only be specified if no other arg is specified. Returns: mapping: New instance of `_Mapping` which has inputs merged with self. Raises: ValueError: if mapping and any other arg is not `None`. """ if mapping is None: mapping = _Mapping(x=x, y=y, ildj_map=ildj_map, kwargs=kwargs) elif any(arg is not None for arg in [x, y, ildj_map, kwargs]): raise ValueError("Cannot simultaneously specify mapping and individual " "arguments.") return _Mapping( x=self._merge(self.x, mapping.x), y=self._merge(self.y, mapping.y), ildj_map=self._merge_dicts(self.ildj_map, mapping.ildj_map), kwargs=self._merge(self.kwargs, mapping.kwargs)) def _merge_dicts(self, old=None, new=None): """Helper to merge two dictionaries.""" old = {} if old is None else old new = {} if new is None else new for k, v in new.items(): val = old.get(k, None) if val is not None and val is not v: raise ValueError("Found different value for existing key " "(key:{} old_value:{} new_value:{}".format( k, old[k], v)) old[k] = v return old def _merge(self, old, new): """Helper to merge which handles merging one value.""" if old is None: return new elif new is not None and old is not new: raise ValueError("Incompatible values: %s != %s" % (old, new)) return old def _deep_tuple(self, x): """Converts lists of lists to tuples of tuples.""" return (tuple(map(self._deep_tuple, x)) if isinstance(x, (list, tuple)) else x) class Bijector(metaclass=abc.ABCMeta): r"""Interface for transformations of a `Distribution` sample. Bijectors can be used to represent any differentiable and injective (one to one) function defined on an open subset of `R^n`. Some non-injective transformations are also supported (see "Non Injective Transforms" below). #### Mathematical Details A `Bijector` implements a [smooth covering map]( https://en.wikipedia.org/wiki/Local_diffeomorphism), i.e., a local diffeomorphism such that every point in the target has a neighborhood evenly covered by a map ([see also]( https://en.wikipedia.org/wiki/Covering_space#Covering_of_a_manifold)). A `Bijector` is used by `TransformedDistribution` but can be generally used for transforming a `Distribution` generated `Tensor`. A `Bijector` is characterized by three operations: 1. Forward Useful for turning one random outcome into another random outcome from a different distribution. 2. Inverse Useful for "reversing" a transformation to compute one probability in terms of another. 3. `log_det_jacobian(x)` "The log of the absolute value of the determinant of the matrix of all first-order partial derivatives of the inverse function." Useful for inverting a transformation to compute one probability in terms of another. Geometrically, the Jacobian determinant is the volume of the transformation and is used to scale the probability. We take the absolute value of the determinant before log to avoid NaN values. Geometrically, a negative determinant corresponds to an orientation-reversing transformation. It is ok for us to discard the sign of the determinant because we only integrate everywhere-nonnegative functions (probability densities) and the correct orientation is always the one that produces a nonnegative integrand. By convention, transformations of random variables are named in terms of the forward transformation. The forward transformation creates samples, the inverse is useful for computing probabilities. #### Example Uses - Basic properties: ```python x = ... # A tensor. # Evaluate forward transformation. fwd_x = my_bijector.forward(x) x == my_bijector.inverse(fwd_x) x != my_bijector.forward(fwd_x) # Not equal because x != g(g(x)). ``` - Computing a log-likelihood: ```python def transformed_log_prob(bijector, log_prob, x): return (bijector.inverse_log_det_jacobian(x, event_ndims=0) + log_prob(bijector.inverse(x))) ``` - Transforming a random outcome: ```python def transformed_sample(bijector, x): return bijector.forward(x) ``` #### Example Bijectors - "Exponential" ```none Y = g(X) = exp(X) X ~ Normal(0, 1) # Univariate. ``` Implies: ```none g^{-1}(Y) = log(Y) |Jacobian(g^{-1})(y)| = 1 / y Y ~ LogNormal(0, 1), i.e., prob(Y=y) = |Jacobian(g^{-1})(y)| * prob(X=g^{-1}(y)) = (1 / y) Normal(log(y); 0, 1) ``` Here is an example of how one might implement the `Exp` bijector: ```python class Exp(Bijector): def __init__(self, validate_args=False, name="exp"): super(Exp, self).__init__( validate_args=validate_args, forward_min_event_ndims=0, name=name) def _forward(self, x): return math_ops.exp(x) def _inverse(self, y): return math_ops.log(y) def _inverse_log_det_jacobian(self, y): return -self._forward_log_det_jacobian(self._inverse(y)) def _forward_log_det_jacobian(self, x): # Notice that we needn't do any reducing, even when`event_ndims > 0`. # The base Bijector class will handle reducing for us; it knows how # to do so because we called `super` `__init__` with # `forward_min_event_ndims = 0`. return x ``` - "Affine" ```none Y = g(X) = sqrtSigma * X + mu X ~ MultivariateNormal(0, I_d) ``` Implies: ```none g^{-1}(Y) = inv(sqrtSigma) * (Y - mu) |Jacobian(g^{-1})(y)| = det(inv(sqrtSigma)) Y ~ MultivariateNormal(mu, sqrtSigma) , i.e., prob(Y=y) = |Jacobian(g^{-1})(y)| * prob(X=g^{-1}(y)) = det(sqrtSigma)^(-d) * MultivariateNormal(inv(sqrtSigma) * (y - mu); 0, I_d) ``` #### Min_event_ndims and Naming Bijectors are named for the dimensionality of data they act on (i.e. without broadcasting). We can think of bijectors having an intrinsic `min_event_ndims` , which is the minimum number of dimensions for the bijector act on. For instance, a Cholesky decomposition requires a matrix, and hence `min_event_ndims=2`. Some examples: `AffineScalar: min_event_ndims=0` `Affine: min_event_ndims=1` `Cholesky: min_event_ndims=2` `Exp: min_event_ndims=0` `Sigmoid: min_event_ndims=0` `SoftmaxCentered: min_event_ndims=1` Note the difference between `Affine` and `AffineScalar`. `AffineScalar` operates on scalar events, whereas `Affine` operates on vector-valued events. More generally, there is a `forward_min_event_ndims` and an `inverse_min_event_ndims`. In most cases, these will be the same. However, for some shape changing bijectors, these will be different (e.g. a bijector which pads an extra dimension at the end, might have `forward_min_event_ndims=0` and `inverse_min_event_ndims=1`. #### Jacobian Determinant The Jacobian determinant is a reduction over `event_ndims - min_event_ndims` (`forward_min_event_ndims` for `forward_log_det_jacobian` and `inverse_min_event_ndims` for `inverse_log_det_jacobian`). To see this, consider the `Exp` `Bijector` applied to a `Tensor` which has sample, batch, and event (S, B, E) shape semantics. Suppose the `Tensor`'s partitioned-shape is `(S=[4], B=[2], E=[3, 3])`. The shape of the `Tensor` returned by `forward` and `inverse` is unchanged, i.e., `[4, 2, 3, 3]`. However the shape returned by `inverse_log_det_jacobian` is `[4, 2]` because the Jacobian determinant is a reduction over the event dimensions. Another example is the `Affine` `Bijector`. Because `min_event_ndims = 1`, the Jacobian determinant reduction is over `event_ndims - 1`. It is sometimes useful to implement the inverse Jacobian determinant as the negative forward Jacobian determinant. For example, ```python def _inverse_log_det_jacobian(self, y): return -self._forward_log_det_jac(self._inverse(y)) # Note negation. ``` The correctness of this approach can be seen from the following claim. - Claim: Assume `Y = g(X)` is a bijection whose derivative exists and is nonzero for its domain, i.e., `dY/dX = d/dX g(X) != 0`. Then: ```none (log o det o jacobian o g^{-1})(Y) = -(log o det o jacobian o g)(X) ``` - Proof: From the bijective, nonzero differentiability of `g`, the [inverse function theorem]( https://en.wikipedia.org/wiki/Inverse_function_theorem) implies `g^{-1}` is differentiable in the image of `g`. Applying the chain rule to `y = g(x) = g(g^{-1}(y))` yields `I = g'(g^{-1}(y))*g^{-1}'(y)`. The same theorem also implies `g^{-1}'` is non-singular therefore: `inv[ g'(g^{-1}(y)) ] = g^{-1}'(y)`. The claim follows from [properties of determinant]( https://en.wikipedia.org/wiki/Determinant#Multiplicativity_and_matrix_groups). Generally its preferable to directly implement the inverse Jacobian determinant. This should have superior numerical stability and will often share subgraphs with the `_inverse` implementation. #### Is_constant_jacobian Certain bijectors will have constant jacobian matrices. For instance, the `Affine` bijector encodes multiplication by a matrix plus a shift, with jacobian matrix, the same aforementioned matrix. `is_constant_jacobian` encodes the fact that the jacobian matrix is constant. The semantics of this argument are the following: * Repeated calls to "log_det_jacobian" functions with the same `event_ndims` (but not necessarily same input), will return the first computed jacobian (because the matrix is constant, and hence is input independent). * `log_det_jacobian` implementations are merely broadcastable to the true `log_det_jacobian` (because, again, the jacobian matrix is input independent). Specifically, `log_det_jacobian` is implemented as the log jacobian determinant for a single input. ```python class Identity(Bijector): def __init__(self, validate_args=False, name="identity"): super(Identity, self).__init__( is_constant_jacobian=True, validate_args=validate_args, forward_min_event_ndims=0, name=name) def _forward(self, x): return x def _inverse(self, y): return y def _inverse_log_det_jacobian(self, y): return -self._forward_log_det_jacobian(self._inverse(y)) def _forward_log_det_jacobian(self, x): # The full log jacobian determinant would be array_ops.zero_like(x). # However, we circumvent materializing that, since the jacobian # calculation is input independent, and we specify it for one input. return constant_op.constant(0., x.dtype.base_dtype) ``` #### Subclass Requirements - Subclasses typically implement: - `_forward`, - `_inverse`, - `_inverse_log_det_jacobian`, - `_forward_log_det_jacobian` (optional). The `_forward_log_det_jacobian` is called when the bijector is inverted via the `Invert` bijector. If undefined, a slightly less efficiently calculation, `-1 * _inverse_log_det_jacobian`, is used. If the bijector changes the shape of the input, you must also implement: - _forward_event_shape_tensor, - _forward_event_shape (optional), - _inverse_event_shape_tensor, - _inverse_event_shape (optional). By default the event-shape is assumed unchanged from input. - If the `Bijector`'s use is limited to `TransformedDistribution` (or friends like `QuantizedDistribution`) then depending on your use, you may not need to implement all of `_forward` and `_inverse` functions. Examples: 1. Sampling (e.g., `sample`) only requires `_forward`. 2. Probability functions (e.g., `prob`, `cdf`, `survival`) only require `_inverse` (and related). 3. Only calling probability functions on the output of `sample` means `_inverse` can be implemented as a cache lookup. See "Example Uses" [above] which shows how these functions are used to transform a distribution. (Note: `_forward` could theoretically be implemented as a cache lookup but this would require controlling the underlying sample generation mechanism.) #### Non Injective Transforms **WARNING** Handing of non-injective transforms is subject to change. Non injective maps `g` are supported, provided their domain `D` can be partitioned into `k` disjoint subsets, `Union{D1, ..., Dk}`, such that, ignoring sets of measure zero, the restriction of `g` to each subset is a differentiable bijection onto `g(D)`. In particular, this implies that for `y in g(D)`, the set inverse, i.e. `g^{-1}(y) = {x in D : g(x) = y}`, always contains exactly `k` distinct points. The property, `_is_injective` is set to `False` to indicate that the bijector is not injective, yet satisfies the above condition. The usual bijector API is modified in the case `_is_injective is False` (see method docstrings for specifics). Here we show by example the `AbsoluteValue` bijector. In this case, the domain `D = (-inf, inf)`, can be partitioned into `D1 = (-inf, 0)`, `D2 = {0}`, and `D3 = (0, inf)`. Let `gi` be the restriction of `g` to `Di`, then both `g1` and `g3` are bijections onto `(0, inf)`, with `g1^{-1}(y) = -y`, and `g3^{-1}(y) = y`. We will use `g1` and `g3` to define bijector methods over `D1` and `D3`. `D2 = {0}` is an oddball in that `g2` is one to one, and the derivative is not well defined. Fortunately, when considering transformations of probability densities (e.g. in `TransformedDistribution`), sets of measure zero have no effect in theory, and only a small effect in 32 or 64 bit precision. For that reason, we define `inverse(0)` and `inverse_log_det_jacobian(0)` both as `[0, 0]`, which is convenient and results in a left-semicontinuous pdf. ```python abs = tfp.distributions.bijectors.AbsoluteValue() abs.forward(-1.) ==> 1. abs.forward(1.) ==> 1. abs.inverse(1.) ==> (-1., 1.) # The |dX/dY| is constant, == 1. So Log|dX/dY| == 0. abs.inverse_log_det_jacobian(1., event_ndims=0) ==> (0., 0.) # Special case handling of 0. abs.inverse(0.) ==> (0., 0.) abs.inverse_log_det_jacobian(0., event_ndims=0) ==> (0., 0.) ``` """ @abc.abstractmethod def __init__(self, graph_parents=None, is_constant_jacobian=False, validate_args=False, dtype=None, forward_min_event_ndims=None, inverse_min_event_ndims=None, name=None): """Constructs Bijector. A `Bijector` transforms random variables into new random variables. Examples: ```python # Create the Y = g(X) = X transform. identity = Identity() # Create the Y = g(X) = exp(X) transform. exp = Exp() ``` See `Bijector` subclass docstring for more details and specific examples. Args: graph_parents: Python list of graph prerequisites of this `Bijector`. is_constant_jacobian: Python `bool` indicating that the Jacobian matrix is not a function of the input. validate_args: Python `bool`, default `False`. Whether to validate input with asserts. If `validate_args` is `False`, and the inputs are invalid, correct behavior is not guaranteed. dtype: `tf.dtype` supported by this `Bijector`. `None` means dtype is not enforced. forward_min_event_ndims: Python `integer` indicating the minimum number of dimensions `forward` operates on. inverse_min_event_ndims: Python `integer` indicating the minimum number of dimensions `inverse` operates on. Will be set to `forward_min_event_ndims` by default, if no value is provided. name: The name to give Ops created by the initializer. Raises: ValueError: If neither `forward_min_event_ndims` and `inverse_min_event_ndims` are specified, or if either of them is negative. ValueError: If a member of `graph_parents` is not a `Tensor`. """ self._graph_parents = graph_parents or [] if forward_min_event_ndims is None and inverse_min_event_ndims is None: raise ValueError("Must specify at least one of `forward_min_event_ndims` " "and `inverse_min_event_ndims`.") elif inverse_min_event_ndims is None: inverse_min_event_ndims = forward_min_event_ndims elif forward_min_event_ndims is None: forward_min_event_ndims = inverse_min_event_ndims if not isinstance(forward_min_event_ndims, int): raise TypeError("Expected forward_min_event_ndims to be of " "type int, got {}".format( type(forward_min_event_ndims).__name__)) if not isinstance(inverse_min_event_ndims, int): raise TypeError("Expected inverse_min_event_ndims to be of " "type int, got {}".format( type(inverse_min_event_ndims).__name__)) if forward_min_event_ndims < 0: raise ValueError("forward_min_event_ndims must be a non-negative " "integer.") if inverse_min_event_ndims < 0: raise ValueError("inverse_min_event_ndims must be a non-negative " "integer.") self._forward_min_event_ndims = forward_min_event_ndims self._inverse_min_event_ndims = inverse_min_event_ndims self._is_constant_jacobian = is_constant_jacobian self._constant_ildj_map = {} self._validate_args = validate_args self._dtype = dtype # These dicts can only be accessed using _Mapping.x_key or _Mapping.y_key self._from_y = {} self._from_x = {} if name: self._name = name else: # We want the default convention to be snake_case rather than CamelCase # since `Chain` uses bijector.name as the kwargs dictionary key. def camel_to_snake(name): s1 = re.sub("(.)([A-Z][a-z]+)", r"\1_\2", name) return re.sub("([a-z0-9])([A-Z])", r"\1_\2", s1).lower() self._name = camel_to_snake(type(self).__name__.lstrip("_")) for i, t in enumerate(self._graph_parents): if t is None or not tensor_util.is_tf_type(t): raise ValueError("Graph parent item %d is not a Tensor; %s." % (i, t)) @property def graph_parents(self): """Returns this `Bijector`'s graph_parents as a Python list.""" return self._graph_parents @property def forward_min_event_ndims(self): """Returns the minimal number of dimensions bijector.forward operates on.""" return self._forward_min_event_ndims @property def inverse_min_event_ndims(self): """Returns the minimal number of dimensions bijector.inverse operates on.""" return self._inverse_min_event_ndims @property def is_constant_jacobian(self): """Returns true iff the Jacobian matrix is not a function of x. Note: Jacobian matrix is either constant for both forward and inverse or neither. Returns: is_constant_jacobian: Python `bool`. """ return self._is_constant_jacobian @property def _is_injective(self): """Returns true iff the forward map `g` is injective (one-to-one function). **WARNING** This hidden property and its behavior are subject to change. Note: Non-injective maps `g` are supported, provided their domain `D` can be partitioned into `k` disjoint subsets, `Union{D1, ..., Dk}`, such that, ignoring sets of measure zero, the restriction of `g` to each subset is a differentiable bijection onto `g(D)`. Returns: is_injective: Python `bool`. """ return True @property def validate_args(self): """Returns True if Tensor arguments will be validated.""" return self._validate_args @property def dtype(self): """dtype of `Tensor`s transformable by this distribution.""" return self._dtype @property def name(self): """Returns the string name of this `Bijector`.""" return self._name def _forward_event_shape_tensor(self, input_shape): """Subclass implementation for `forward_event_shape_tensor` function.""" # By default, we assume event_shape is unchanged. return input_shape def forward_event_shape_tensor(self, input_shape, name="forward_event_shape_tensor"): """Shape of a single sample from a single batch as an `int32` 1D `Tensor`. Args: input_shape: `Tensor`, `int32` vector indicating event-portion shape passed into `forward` function. name: name to give to the op Returns: forward_event_shape_tensor: `Tensor`, `int32` vector indicating event-portion shape after applying `forward`. """ with self._name_scope(name, [input_shape]): input_shape = ops.convert_to_tensor(input_shape, dtype=dtypes.int32, name="input_shape") return self._forward_event_shape_tensor(input_shape) def _forward_event_shape(self, input_shape): """Subclass implementation for `forward_event_shape` public function.""" # By default, we assume event_shape is unchanged. return input_shape def forward_event_shape(self, input_shape): """Shape of a single sample from a single batch as a `TensorShape`. Same meaning as `forward_event_shape_tensor`. May be only partially defined. Args: input_shape: `TensorShape` indicating event-portion shape passed into `forward` function. Returns: forward_event_shape_tensor: `TensorShape` indicating event-portion shape after applying `forward`. Possibly unknown. """ return self._forward_event_shape(tensor_shape.TensorShape(input_shape)) def _inverse_event_shape_tensor(self, output_shape): """Subclass implementation for `inverse_event_shape_tensor` function.""" # By default, we assume event_shape is unchanged. return output_shape def inverse_event_shape_tensor(self, output_shape, name="inverse_event_shape_tensor"): """Shape of a single sample from a single batch as an `int32` 1D `Tensor`. Args: output_shape: `Tensor`, `int32` vector indicating event-portion shape passed into `inverse` function. name: name to give to the op Returns: inverse_event_shape_tensor: `Tensor`, `int32` vector indicating event-portion shape after applying `inverse`. """ with self._name_scope(name, [output_shape]): output_shape = ops.convert_to_tensor(output_shape, dtype=dtypes.int32, name="output_shape") return self._inverse_event_shape_tensor(output_shape) def _inverse_event_shape(self, output_shape): """Subclass implementation for `inverse_event_shape` public function.""" # By default, we assume event_shape is unchanged. return tensor_shape.TensorShape(output_shape) def inverse_event_shape(self, output_shape): """Shape of a single sample from a single batch as a `TensorShape`. Same meaning as `inverse_event_shape_tensor`. May be only partially defined. Args: output_shape: `TensorShape` indicating event-portion shape passed into `inverse` function. Returns: inverse_event_shape_tensor: `TensorShape` indicating event-portion shape after applying `inverse`. Possibly unknown. """ return self._inverse_event_shape(output_shape) def _forward(self, x): """Subclass implementation for `forward` public function.""" raise NotImplementedError("forward not implemented.") def _call_forward(self, x, name, **kwargs): with self._name_scope(name, [x]): x = ops.convert_to_tensor(x, name="x") self._maybe_assert_dtype(x) if not self._is_injective: # No caching for non-injective return self._forward(x, **kwargs) mapping = self._lookup(x=x, kwargs=kwargs) if mapping.y is not None: return mapping.y mapping = mapping.merge(y=self._forward(x, **kwargs)) self._cache(mapping) return mapping.y def forward(self, x, name="forward"): """Returns the forward `Bijector` evaluation, i.e., X = g(Y). Args: x: `Tensor`. The input to the "forward" evaluation. name: The name to give this op. Returns: `Tensor`. Raises: TypeError: if `self.dtype` is specified and `x.dtype` is not `self.dtype`. NotImplementedError: if `_forward` is not implemented. """ return self._call_forward(x, name) def _inverse(self, y): """Subclass implementation for `inverse` public function.""" raise NotImplementedError("inverse not implemented") def _call_inverse(self, y, name, **kwargs): with self._name_scope(name, [y]): y = ops.convert_to_tensor(y, name="y") self._maybe_assert_dtype(y) if not self._is_injective: # No caching for non-injective return self._inverse(y, **kwargs) mapping = self._lookup(y=y, kwargs=kwargs) if mapping.x is not None: return mapping.x mapping = mapping.merge(x=self._inverse(y, **kwargs)) self._cache(mapping) return mapping.x def inverse(self, y, name="inverse"): """Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y). Args: y: `Tensor`. The input to the "inverse" evaluation. name: The name to give this op. Returns: `Tensor`, if this bijector is injective. If not injective, returns the k-tuple containing the unique `k` points `(x1, ..., xk)` such that `g(xi) = y`. Raises: TypeError: if `self.dtype` is specified and `y.dtype` is not `self.dtype`. NotImplementedError: if `_inverse` is not implemented. """ return self._call_inverse(y, name) def _inverse_log_det_jacobian(self, y): """Subclass implementation of `inverse_log_det_jacobian` public function. In particular, this method differs from the public function, in that it does not take `event_ndims`. Thus, this implements the minimal Jacobian determinant calculation (i.e. over `inverse_min_event_ndims`). Args: y: `Tensor`. The input to the "inverse_log_det_jacobian" evaluation. Returns: inverse_log_det_jacobian: `Tensor`, if this bijector is injective. If not injective, returns the k-tuple containing jacobians for the unique `k` points `(x1, ..., xk)` such that `g(xi) = y`. """ raise NotImplementedError("inverse_log_det_jacobian not implemented.") def _call_inverse_log_det_jacobian(self, y, event_ndims, name, **kwargs): with self._name_scope(name, [y]): if event_ndims in self._constant_ildj_map: return self._constant_ildj_map[event_ndims] y = ops.convert_to_tensor(y, name="y") self._maybe_assert_dtype(y) with ops.control_dependencies(self._check_valid_event_ndims( min_event_ndims=self.inverse_min_event_ndims, event_ndims=event_ndims)): if not self._is_injective: # No caching for non-injective try: ildjs = self._inverse_log_det_jacobian(y, **kwargs) return tuple(self._reduce_jacobian_det_over_event( y, ildj, self.inverse_min_event_ndims, event_ndims) for ildj in ildjs) except NotImplementedError as original_exception: try: x = self._inverse(y, **kwargs) fldjs = self._forward_log_det_jacobian(x, **kwargs) return tuple(self._reduce_jacobian_det_over_event( x, -fldj, self.forward_min_event_ndims, event_ndims) for fldj in fldjs) except NotImplementedError: raise original_exception mapping = self._lookup(y=y, kwargs=kwargs) if mapping.ildj_map is not None and event_ndims in mapping.ildj_map: return mapping.ildj_map[event_ndims] try: x = None # Not needed; leave cache as is. ildj = self._inverse_log_det_jacobian(y, **kwargs) ildj = self._reduce_jacobian_det_over_event( y, ildj, self.inverse_min_event_ndims, event_ndims) except NotImplementedError as original_exception: try: x = (mapping.x if mapping.x is not None else self._inverse(y, **kwargs)) ildj = -self._forward_log_det_jacobian(x, **kwargs) ildj = self._reduce_jacobian_det_over_event( x, ildj, self.forward_min_event_ndims, event_ndims) except NotImplementedError: raise original_exception mapping = mapping.merge(x=x, ildj_map={event_ndims: ildj}) self._cache(mapping) if self.is_constant_jacobian: self._constant_ildj_map[event_ndims] = ildj return ildj def inverse_log_det_jacobian( self, y, event_ndims, name="inverse_log_det_jacobian"): """Returns the (log o det o Jacobian o inverse)(y). Mathematically, returns: `log(det(dX/dY))(Y)`. (Recall that: `X=g^{-1}(Y)`.) Note that `forward_log_det_jacobian` is the negative of this function, evaluated at `g^{-1}(y)`. Args: y: `Tensor`. The input to the "inverse" Jacobian determinant evaluation. event_ndims: Number of dimensions in the probabilistic events being transformed. Must be greater than or equal to `self.inverse_min_event_ndims`. The result is summed over the final dimensions to produce a scalar Jacobian determinant for each event, i.e. it has shape `y.shape.ndims - event_ndims` dimensions. name: The name to give this op. Returns: `Tensor`, if this bijector is injective. If not injective, returns the tuple of local log det Jacobians, `log(det(Dg_i^{-1}(y)))`, where `g_i` is the restriction of `g` to the `ith` partition `Di`. Raises: TypeError: if `self.dtype` is specified and `y.dtype` is not `self.dtype`. NotImplementedError: if `_inverse_log_det_jacobian` is not implemented. """ return self._call_inverse_log_det_jacobian(y, event_ndims, name) def _forward_log_det_jacobian(self, x): """Subclass implementation of `forward_log_det_jacobian` public function. In particular, this method differs from the public function, in that it does not take `event_ndims`. Thus, this implements the minimal Jacobian determinant calculation (i.e. over `forward_min_event_ndims`). Args: x: `Tensor`. The input to the "forward_log_det_jacobian" evaluation. Returns: forward_log_det_jacobian: `Tensor`, if this bijector is injective. If not injective, returns the k-tuple containing jacobians for the unique `k` points `(x1, ..., xk)` such that `g(xi) = y`. """ raise NotImplementedError( "forward_log_det_jacobian not implemented.") def _call_forward_log_det_jacobian(self, x, event_ndims, name, **kwargs): if not self._is_injective: raise NotImplementedError( "forward_log_det_jacobian cannot be implemented for non-injective " "transforms.") with self._name_scope(name, [x]): with ops.control_dependencies(self._check_valid_event_ndims( min_event_ndims=self.forward_min_event_ndims, event_ndims=event_ndims)): if event_ndims in self._constant_ildj_map: # Need "-1. *" to avoid invalid-unary-operand-type linter warning. return -1. * self._constant_ildj_map[event_ndims] x = ops.convert_to_tensor(x, name="x") self._maybe_assert_dtype(x) if not self._is_injective: # No caching for non-injective try: fldjs = self._forward_log_det_jacobian(x, **kwargs) # No caching. return tuple(self._reduce_jacobian_det_over_event( x, fldj, self.forward_min_event_ndims, event_ndims) for fldj in fldjs) except NotImplementedError as original_exception: try: y = self._forward(x, **kwargs) ildjs = self._inverse_log_det_jacobian(y, **kwargs) return tuple(self._reduce_jacobian_det_over_event( y, -ildj, self.inverse_min_event_ndims, event_ndims) for ildj in ildjs) except NotImplementedError: raise original_exception mapping = self._lookup(x=x, kwargs=kwargs) if mapping.ildj_map is not None and event_ndims in mapping.ildj_map: return -mapping.ildj_map[event_ndims] try: y = None # Not needed; leave cache as is. ildj = -self._forward_log_det_jacobian(x, **kwargs) ildj = self._reduce_jacobian_det_over_event( x, ildj, self.forward_min_event_ndims, event_ndims) except NotImplementedError as original_exception: try: y = (mapping.y if mapping.y is not None else self._forward(x, **kwargs)) ildj = self._inverse_log_det_jacobian(y, **kwargs) ildj = self._reduce_jacobian_det_over_event( y, ildj, self.inverse_min_event_ndims, event_ndims) except NotImplementedError: raise original_exception mapping = mapping.merge(y=y, ildj_map={event_ndims: ildj}) self._cache(mapping) if self.is_constant_jacobian: self._constant_ildj_map[event_ndims] = ildj return -ildj def forward_log_det_jacobian( self, x, event_ndims, name="forward_log_det_jacobian"): """Returns both the forward_log_det_jacobian. Args: x: `Tensor`. The input to the "forward" Jacobian determinant evaluation. event_ndims: Number of dimensions in the probabilistic events being transformed. Must be greater than or equal to `self.forward_min_event_ndims`. The result is summed over the final dimensions to produce a scalar Jacobian determinant for each event, i.e. it has shape `x.shape.ndims - event_ndims` dimensions. name: The name to give this op. Returns: `Tensor`, if this bijector is injective. If not injective this is not implemented. Raises: TypeError: if `self.dtype` is specified and `y.dtype` is not `self.dtype`. NotImplementedError: if neither `_forward_log_det_jacobian` nor {`_inverse`, `_inverse_log_det_jacobian`} are implemented, or this is a non-injective bijector. """ return self._call_forward_log_det_jacobian(x, event_ndims, name) @contextlib.contextmanager def _name_scope(self, name=None, values=None): """Helper function to standardize op scope.""" with ops.name_scope(self.name): with ops.name_scope( name, values=(values or []) + self.graph_parents) as scope: yield scope def _maybe_assert_dtype(self, x): """Helper to check dtype when self.dtype is known.""" if self.dtype is not None and self.dtype.base_dtype != x.dtype.base_dtype: raise TypeError("Input had dtype %s but expected %s." % (self.dtype, x.dtype)) def _cache(self, mapping): """Helper which stores mapping info in forward/inverse dicts.""" # Merging from lookup is an added check that we're not overwriting anything # which is not None. mapping = mapping.merge(mapping=self._lookup( mapping.x, mapping.y, mapping.kwargs)) if mapping.x is None and mapping.y is None: raise ValueError("Caching expects at least one of (x,y) to be known, " "i.e., not None.") self._from_x[mapping.x_key] = mapping self._from_y[mapping.y_key] = mapping def _lookup(self, x=None, y=None, kwargs=None): """Helper which retrieves mapping info from forward/inverse dicts.""" mapping = _Mapping(x=x, y=y, kwargs=kwargs) # Since _cache requires both x,y to be set, we only need to do one cache # lookup since the mapping is always in both or neither. if mapping.x is not None: return self._from_x.get(mapping.x_key, mapping) if mapping.y is not None: return self._from_y.get(mapping.y_key, mapping) return mapping def _reduce_jacobian_det_over_event( self, y, ildj, min_event_ndims, event_ndims): """Reduce jacobian over event_ndims - min_event_ndims.""" # In this case, we need to tile the Jacobian over the event and reduce. y_rank = array_ops.rank(y) y_shape = array_ops.shape(y)[ y_rank - event_ndims : y_rank - min_event_ndims] ones = array_ops.ones(y_shape, ildj.dtype) reduced_ildj = math_ops.reduce_sum( ones * ildj, axis=self._get_event_reduce_dims(min_event_ndims, event_ndims)) # The multiplication by ones can change the inferred static shape so we try # to recover as much as possible. event_ndims_ = self._maybe_get_static_event_ndims(event_ndims) if (event_ndims_ is not None and y.shape.ndims is not None and ildj.shape.ndims is not None): y_shape = y.shape[y.shape.ndims - event_ndims_ : y.shape.ndims - min_event_ndims] broadcast_shape = array_ops.broadcast_static_shape(ildj.shape, y_shape) reduced_ildj.set_shape( broadcast_shape[: broadcast_shape.ndims - ( event_ndims_ - min_event_ndims)]) return reduced_ildj def _get_event_reduce_dims(self, min_event_ndims, event_ndims): """Compute the reduction dimensions given event_ndims.""" event_ndims_ = self._maybe_get_static_event_ndims(event_ndims) if event_ndims_ is not None: return [-index for index in range(1, event_ndims_ - min_event_ndims + 1)] else: reduce_ndims = event_ndims - min_event_ndims return math_ops.range(-reduce_ndims, 0) def _check_valid_event_ndims(self, min_event_ndims, event_ndims): """Check whether event_ndims is at least min_event_ndims.""" event_ndims = ops.convert_to_tensor(event_ndims, name="event_ndims") event_ndims_ = tensor_util.constant_value(event_ndims) assertions = [] if not event_ndims.dtype.is_integer: raise ValueError("Expected integer dtype, got dtype {}".format( event_ndims.dtype)) if event_ndims_ is not None: if event_ndims.shape.ndims != 0: raise ValueError("Expected scalar event_ndims, got shape {}".format( event_ndims.shape)) if min_event_ndims > event_ndims_: raise ValueError("event_ndims ({}) must be larger than " "min_event_ndims ({})".format( event_ndims_, min_event_ndims)) elif self.validate_args: assertions += [ check_ops.assert_greater_equal(event_ndims, min_event_ndims)] if event_ndims.shape.is_fully_defined(): if event_ndims.shape.ndims != 0: raise ValueError("Expected scalar shape, got ndims {}".format( event_ndims.shape.ndims)) elif self.validate_args: assertions += [ check_ops.assert_rank(event_ndims, 0, message="Expected scalar.")] return assertions def _maybe_get_static_event_ndims(self, event_ndims): """Helper which returns tries to return an integer static value.""" event_ndims_ = distribution_util.maybe_get_static_value(event_ndims) if isinstance(event_ndims_, (np.generic, np.ndarray)): if event_ndims_.dtype not in (np.int32, np.int64): raise ValueError("Expected integer dtype, got dtype {}".format( event_ndims_.dtype)) if isinstance(event_ndims_, np.ndarray) and len(event_ndims_.shape): raise ValueError("Expected a scalar integer, got {}".format( event_ndims_)) event_ndims_ = int(event_ndims_) return event_ndims_