tfp.substrates.jax.bijectors.Bijector

Interface for transformations of a Distribution sample.

View aliases

Main aliases

tfp.experimental.substrates.jax.bijectors.Bijector

@abc.abstractmethod
tfp.substrates.jax.bijectors.Bijector(
    graph_parents=None, is_constant_jacobian=False, validate_args=False, dtype=None,
    forward_min_event_ndims=UNSPECIFIED, inverse_min_event_ndims=UNSPECIFIED,
    parameters=None, name=None
)

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, i.e., a local diffeomorphism such that every point in the target has a neighborhood evenly covered by a map (see also). A Bijector is used by TransformedDistribution but can be generally used for transforming a Distribution generated Tensor. A Bijector is characterized by three operations:

Forward

Useful for turning one random outcome into another random outcome from a different distribution.
Inverse

Useful for 'reversing' a transformation to compute one probability in terms of another.
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:

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:

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:

def transformed_sample(bijector, x):
  return bijector.forward(x)

Example Bijectors

'Exponential'

Y = g(X) = exp(X)
X ~ Normal(0, 1)  # Univariate.

Implies:

  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:

  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 tf.exp(x)

    def _inverse(self, y):
      return tf.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'

Y = g(X) = sqrtSigma * X + mu
X ~ MultivariateNormal(0, I_d)

Implies:

  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.

Bijectors which operate on structures of Tensor require structured min_event_ndims matching the structure of the inputs. In these cases, min_event_ndims describes both the minimum dimensionality and the structure of arguments to forward and inverse. For example:

Split([sizes], axis):
  forward_min_event_ndims=-axis
  inverse_min_event_ndims=[-axis] * len(sizes)

In these cases, the leftmost min_event_ndims[i] elements of tensor[i].shape must be identical for all structured inputs i.

Finally, some bijectors that operate on structures of inputs may not know the minimum structured rank of their inputs without calltime shape information (Composite bijectors, for example). In these cases, both min_event_ndims properties will indicate the expected structure of inputs and outputs, but the component values may be None.

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,

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:

(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 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.

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.

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 tf.zero_like(x).
    # However, we circumvent materializing that, since the jacobian
    # calculation is input independent, and we specify it for one input.
    return tf.constant(0., x.dtype)

Subclass Requirements

Subclasses typically implement:
- _forward,
- _inverse,
- _inverse_log_det_jacobian,
- _forward_log_det_jacobian (optional),
- _is_increasing (scalar bijectors only)
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.

Multipart bijectors, which operate on structures of tensors, may implement additional methods to propogate calltime dtype information over any changes to structure. These methods are:
- _forward_dtype
- _inverse_dtype
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

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.

abs = tfp.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.)

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. For multipart bijectors, this value is expected to be the same for all elements of the input and output structures.
`forward_min_event_ndims`	Python `integer` (structure) indicating the minimum number of dimensions on which `forward` operates.
`inverse_min_event_ndims`	Python `integer` (structure) indicating the minimum number of dimensions on which `inverse` operates. Will be set to `forward_min_event_ndims` by default, if no value is provided.
`parameters`	Python `dict` of parameters used to instantiate this `Bijector`. Bijector instances with identical types, names, and `parameters` share an input/output cache. `parameters` dicts are keyed by strings and are identical if their keys are identical and if corresponding values have identical hashes (or object ids, for unhashable objects).
`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`.

Attributes
`dtype`
`forward_min_event_ndims`	Returns the minimal number of dimensions bijector.forward operates on. Multipart bijectors return structured `ndims`, which indicates the expected structure of their inputs. Some multipart bijectors, notably Composites, may return structures of `None`.
`graph_parents`	Returns this `Bijector`'s graph_parents as a Python list.
`inverse_min_event_ndims`	Returns the minimal number of dimensions bijector.inverse operates on. Multipart bijectors return structured `event_ndims`, which indicates the expected structure of their outputs. Some multipart bijectors, notably Composites, may return structures of `None`.
`is_constant_jacobian`	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.
`name`	Returns the string name of this `Bijector`.
`parameters`	Dictionary of parameters used to instantiate this `Bijector`.
`trainable_variables`
`validate_args`	Returns True if Tensor arguments will be validated.
`variables`

Args
`x`	`Tensor` (structure). The input to the 'forward' evaluation.
`name`	The name to give this op.
`**kwargs`	Named arguments forwarded to subclass implementation.

Raises
`TypeError`	if `self.dtype` is specified and `x.dtype` is not `self.dtype`.
`NotImplementedError`	if `_forward` is not implemented.

Args
`input_shape`	`Tensor`, `int32` vector (structure) indicating event-portion shape passed into `forward` function.
`name`	name to give to the op

Args
`x`	`Tensor` (structure). 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 `rank(x) - event_ndims` dimensions. Multipart bijectors require structured event_ndims, such that `rank(y[i]) - rank(event_ndims[i])` is the same for all elements `i` of the structured input. Furthermore, the first `event_ndims[i]` of each `x[i].shape` must be the same for all `i` (broadcasting is not allowed).
`name`	The name to give this op.
`**kwargs`	Named arguments forwarded to subclass implementation.

Raises
`TypeError`	if `y`'s dtype is incompatible with the expected output dtype.
`NotImplementedError`	if neither `_forward_log_det_jacobian` nor {`_inverse`, `_inverse_log_det_jacobian`} are implemented, or this is a non-injective bijector.

Raises
`TypeError`	if `y`'s structured dtype is incompatible with the expected output dtype.
`NotImplementedError`	if `_inverse` is not implemented.

Args
`output_shape`	`Tensor`, `int32` vector (structure) indicating event-portion shape passed into `inverse` function.
`name`	name to give to the op

Raises
`TypeError`	if `x`'s dtype is incompatible with the expected inverse-dtype.
`NotImplementedError`	if `_inverse_log_det_jacobian` is not implemented.

Args
`value`	A `tfd.Distribution`, `tfb.Bijector`, or a (structure of) `Tensor`.
`name`	Python `str` name given to ops created by this function.
`**kwargs`	Additional keyword arguments passed into the created `tfd.TransformedDistribution`, `tfb.Bijector`, or `self.forward`.

tfp.substrates.jax.bijectors.Bijector

View aliases

Mathematical Details

Example Uses

Example Bijectors

Min_event_ndims and Naming

Some examples:

Jacobian Determinant

Is_constant_jacobian

Subclass Requirements

Non Injective Transforms

Args

Raises

Attributes

Methods

forward

forward_dtype

forward_event_ndims

forward_event_shape

forward_event_shape_tensor

forward_log_det_jacobian

inverse

inverse_dtype

inverse_event_ndims

inverse_event_shape

inverse_event_shape_tensor

inverse_log_det_jacobian

__call__

Examples

__eq__

__ne__

`forward`

`forward_dtype`

`forward_event_ndims`

`forward_event_shape`

`forward_event_shape_tensor`

`forward_log_det_jacobian`

`inverse`

`inverse_dtype`

`inverse_event_ndims`

`inverse_event_shape`

`inverse_event_shape_tensor`

`inverse_log_det_jacobian`

`call`

`eq`

`ne`