oryx.bijectors.RationalQuadraticSpline

A piecewise rational quadratic spline, as developed in [1].

Inherits From: Bijector

This transformation represents a monotonically increasing piecewise rational quadratic function. Outside of the bounds of knot_x/knot_y, the transform behaves as an identity function.

Typically this bijector will be used as part of a chain, with splines for trailing x dimensions conditioned on some of the earlier x dimensions, and with the inverse then solved first for unconditioned dimensions, then using conditioning derived from those inverses, and so forth. For example, if we split a 15-D xs vector into 3 components, we may implement a forward and inverse as follows:

nsplits = 3

class SplineParams(tf.Module):

  def __init__(self, nbins=32):
    self._nbins = nbins
    self._built = False
    self._bin_widths = None
    self._bin_heights = None
    self._knot_slopes = None

  def __call__(self, x, nunits):
    if not self._built:
      def _bin_positions(x):
        out_shape = tf.concat((tf.shape(x)[:-1], (nunits, self._nbins)), 0)
        x = tf.reshape(x, out_shape)
        return tf.math.softmax(x, axis=-1) * (2 - self._nbins * 1e-2) + 1e-2

      def _slopes(x):
        out_shape = tf.concat((
          tf.shape(x)[:-1], (nunits, self._nbins - 1)), 0)
        x = tf.reshape(x, out_shape)
        return tf.math.softplus(x) + 1e-2

      self._bin_widths = tf.keras.layers.Dense(
          nunits * self._nbins, activation=_bin_positions, name='w')
      self._bin_heights = tf.keras.layers.Dense(
          nunits * self._nbins, activation=_bin_positions, name='h')
      self._knot_slopes = tf.keras.layers.Dense(
          nunits * (self._nbins - 1), activation=_slopes, name='s')
      self._built = True

    return tfb.RationalQuadraticSpline(
        bin_widths=self._bin_widths(x),
        bin_heights=self._bin_heights(x),
        knot_slopes=self._knot_slopes(x))

xs = np.random.randn(3, 15).astype(np.float32)  # Keras won't Dense(.)(vec).
splines = [SplineParams() for _ in range(nsplits)]

def spline_flow():
  stack = tfb.Identity()
  for i in range(nsplits):
    stack = tfb.RealNVP(5 * i, bijector_fn=splines[i])(stack)
  return stack

ys = spline_flow().forward(xs)
ys_inv = spline_flow().inverse(ys)  # ys_inv ~= xs

For a one-at-a-time autoregressive flow as in [1], it would be profitable to implement a mask over xs to parallelize either the inverse or the forward pass and implement the other using a tf.while_loop. See tfp.bijectors.MaskedAutoregressiveFlow for support doing so (paired with tfp.bijectors.Invert depending which direction should be parallel).

References

[1]: Conor Durkan, Artur Bekasov, Iain Murray, George Papamakarios. Neural Spline Flows. arXiv preprint arXiv:1906.04032, 2019. https://arxiv.org/abs/1906.04032

bin_widths The widths of the spans between subsequent knot x positions, a floating point Tensor. Must be positive, and at least 1-D. Innermost axis must sum to the same value as bin_heights. The knot x positions will be a first at range_min, followed by knots at range_min + cumsum(bin_widths, axis=-1).
bin_heights The heights of the spans between subsequent knot y positions, a floating point Tensor. Must be positive, and at least 1-D. Innermost axis must sum to the same value as bin_widths. The knot y positions will be a first at range_min, followed by knots at range_min + cumsum(bin_heights, axis=-1).
knot_slopes The slope of the spline at each knot, a floating point Tensor. Must be positive. 1s are implicitly padded for the first and last implicit knots corresponding to range_min and range_min + sum(bin_widths, axis=-1). Innermost axis size should be 1 less than that of bin_widths/bin_heights, or 1 for broadcasting.
range_min The x/y position of the first knot, which has implicit slope 1. range_max is implicit, and can be computed as range_min + sum(bin_widths, axis=-1). Scalar floating point Tensor.
validate_args Toggles argument validation (can hurt performance).
name Optional name scope for associated ops. (Defaults to 'RationalQuadraticSpline').

bin_heights

bin_widths

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.
has_static_min_event_ndims Returns True if the bijector has statically-known min_event_ndims.
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.

knot_slopes

name Returns the string name of this Bijector.
parameters Dictionary of parameters used to instantiate this Bijector.
range_min

trainable_variables

validate_args Returns True if Tensor arguments will be validated.
variables

Methods

copy

Creates a copy of the bijector.

Args
**override_parameters_kwargs String/value dictionary of initialization arguments to override with new values.

Returns
bijector A new instance of type(self) initialized from the union of self.parameters and override_parameters_kwargs, i.e., dict(self.parameters, **override_parameters_kwargs).

experimental_batch_shape

Returns the batch shape of this bijector for inputs of the given rank.

The batch shape of a bijector decribes the set of distinct transformations it represents on events of a given size. For example: the bijector tfb.Scale([1., 2.]) has batch shape [2] for scalar events (event_ndims = 0), because applying it to a scalar event produces two scalar outputs, the result of two different scaling transformations. The same bijector has batch shape [] for vector events, because applying it to a vector produces (via elementwise multiplication) a single vector output.

Bijectors that operate independently on multiple state parts, such as tfb.JointMap, must broadcast to a coherent batch shape. Some events may not be valid: for example, the bijector tfd.JointMap([tfb.Scale([1., 2.]), tfb.Scale([1., 2., 3.])]) does not produce a valid batch shape when event_ndims = [0, 0], since the batch shapes of the two parts are inconsistent. The same bijector does define valid batch shapes of [], [2], and [3] if event_ndims is [1, 1], [0, 1], or [1, 0], respectively.

Since transforming a single event produces a scalar log-det-Jacobian, the batch shape of a bijector with non-constant Jacobian is expected to equal the shape of forward_log_det_jacobian(x, event_ndims=x_event_ndims) or inverse_log_det_jacobian(y, event_ndims=y_event_ndims), for x or y of the specified ndims.

Args
x_event_ndims Optional Python int (structure) number of dimensions in a probabilistic event passed to forward; this must be greater than or equal to self.forward_min_event_ndims. If None, defaults to self.forward_min_event_ndims. Mutually exclusive with y_event_ndims. Default value: None.
y_event_ndims Optional Python int (structure) number of dimensions in a probabilistic event passed to inverse; this must be greater than or equal to self.inverse_min_event_ndims. Mutually exclusive with x_event_ndims. Default value: None.

Returns
batch_shape TensorShape batch shape of this bijector for a value with the given event rank. May be unknown or partially defined.

experimental_batch_shape_tensor

Returns the batch shape of this bijector for inputs of the given rank.

The batch shape of a bijector decribes the set of distinct transformations it represents on events of a given size. For example: the bijector tfb.Scale([1., 2.]) has batch shape [2] for scalar events (event_ndims = 0), because applying it to a scalar event produces two scalar outputs, the result of two different scaling transformations. The same bijector has batch shape [] for vector events, because applying it to a vector produces (via elementwise multiplication) a single vector output.

Bijectors that operate independently on multiple state parts, such as tfb.JointMap, must broadcast to a coherent batch shape. Some events may not be valid: for example, the bijector tfd.JointMap([tfb.Scale([1., 2.]), tfb.Scale([1., 2., 3.])]) does not produce a valid batch shape when event_ndims = [0, 0], since the batch shapes of the two parts are inconsistent. The same bijector does define valid batch shapes of [], [2], and [3] if event_ndims is [1, 1], [0, 1], or [1, 0], respectively.

Since transforming a single event produces a scalar log-det-Jacobian, the batch shape of a bijector with non-constant Jacobian is expected to equal the shape of forward_log_det_jacobian(x, event_ndims=x_event_ndims) or inverse_log_det_jacobian(y, event_ndims=y_event_ndims), for x or y of the specified ndims.

Args
x_event_ndims Optional Python int (structure) number of dimensions in a probabilistic event passed to forward; this must be greater than or equal to self.forward_min_event_ndims. If None, defaults to self.forward_min_event_ndims. Mutually exclusive with y_event_ndims. Default value: None.
y_event_ndims Optional Python int (structure) number of dimensions in a probabilistic event passed to inverse; this must be greater than or equal to self.inverse_min_event_ndims. Mutually exclusive with x_event_ndims. Default value: None.

Returns
batch_shape_tensor integer Tensor batch shape of this bijector for a value with the given event rank.

forward

Returns the forward Bijector evaluation, i.e., X = g(Y).

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

Returns
Tensor (structure).

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

forward_dtype

Returns the dtype returned by forward for the provided input.

forward_event_ndims

Returns the number of event dimensions produced by forward.

forward_event_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 (structure) indicating event-portion shape passed into forward function.

Returns
forward_event_shape_tensor TensorShape (structure) indicating event-portion shape after applying forward. Possibly unknown.

forward_event_shape_tensor

Shape of a single sample from a single batch as an int32 1D Tensor.

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

Returns
forward_event_shape_tensor Tensor, int32 vector (structure) indicating event-portion shape after applying forward.

forward_log_det_jacobian

Returns both the forward_log_det_jacobian.

Args
x Tensor (structure). The input to the 'forward' Jacobian determinant evaluation.
event_ndims Optional number of dimensions in the probabilistic events being transformed; this must be greater than or equal to self.forward_min_event_ndims. If event_ndims is specified, the log Jacobian determinant is summed to produce a scalar log-determinant for each event. Otherwise (if event_ndims is None), no reduction is performed. Multipart bijectors require structured event_ndims, such that the batch rank rank(y[i]) - event_ndims[i] is the same for all elements i of the structured input. In most cases (with the exception of tfb.JointMap) they further require that event_ndims[i] - self.inverse_min_event_ndims[i] is the same for all elements i of the structured input. Default value: None (equivalent to self.forward_min_event_ndims).
name The name to give this op.
**kwargs Named arguments forwarded to subclass implementation.

Returns
Tensor (structure), if this bijector is injective. If not injective this is not implemented.

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.
ValueError if the value of event_ndims is not valid for this bijector.

inverse

Returns the inverse Bijector evaluation, i.e., X = g^{-1}(Y).

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

Returns
Tensor (structure), 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 y's structured dtype is incompatible with the expected output dtype.
NotImplementedError if _inverse is not implemented.

inverse_dtype

Returns the dtype returned by inverse for the provided input.

inverse_event_ndims

Returns the number of event dimensions produced by inverse.

inverse_event_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 (structure) indicating event-portion shape passed into inverse function.

Returns
inverse_event_shape_tensor TensorShape (structure) indicating event-portion shape after applying inverse. Possibly unknown.

inverse_event_shape_tensor

Shape of a single sample from a single batch as an int32 1D Tensor.

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

Returns
inverse_event_shape_tensor Tensor, int32 vector (structure) indicating event-portion shape after applying inverse.

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 (structure). The input to the 'inverse' Jacobian determinant evaluation.
event_ndims Optional number of dimensions in the probabilistic events being transformed; this must be greater than or equal to self.inverse_min_event_ndims. If event_ndims is specified, the log Jacobian determinant is summed to produce a scalar log-determinant for each event. Otherwise (if event_ndims is None), no reduction is performed. Multipart bijectors require structured event_ndims, such that the batch rank rank(y[i]) - event_ndims[i] is the same for all elements i of the structured input. In most cases (with the exception of tfb.JointMap) they further require that event_ndims[i] - self.inverse_min_event_ndims[i] is the same for all elements i of the structured input. Default value: None (equivalent to self.inverse_min_event_ndims).
name The name to give this op.
**kwargs Named arguments forwarded to subclass implementation.

Returns
ildj 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 x's dtype is incompatible with the expected inverse-dtype.
NotImplementedError if _inverse_log_det_jacobian is not implemented.
ValueError if the value of event_ndims is not valid for this bijector.

parameter_properties

Returns a dict mapping constructor arg names to property annotations.

This dict should include an entry for each of the bijector's Tensor-valued constructor arguments.

Args
dtype Optional float dtype to assume for continuous-valued parameters. Some constraining bijectors require advance knowledge of the dtype because certain constants (e.g., tfb.Softplus.low) must be instantiated with the same dtype as the values to be transformed.

Returns
parameter_properties A str ->tfp.python.internal.parameter_properties.ParameterPropertiesdict mapping constructor argument names toParameterProperties` instances.

__call__

Applies or composes the Bijector, depending on input type.

This is a convenience function which applies the Bijector instance in three different ways, depending on the input:

  1. If the input is a tfd.Distribution instance, return tfd.TransformedDistribution(distribution=input, bijector=self).
  2. If the input is a tfb.Bijector instance, return tfb.Chain([self, input]).
  3. Otherwise, return self.forward(input)

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.

Returns
composition A tfd.TransformedDistribution if the input was a tfd.Distribution, a tfb.Chain if the input was a tfb.Bijector, or a (structure of) Tensor computed by self.forward.

Examples

sigmoid = tfb.Reciprocal()(
    tfb.Shift(shift=1.)(
      tfb.Exp()(
        tfb.Scale(scale=-1.))))
# ==> `tfb.Chain([
#         tfb.Reciprocal(),
#         tfb.Shift(shift=1.),
#         tfb.Exp(),
#         tfb.Scale(scale=-1.),
#      ])`  # ie, `tfb.Sigmoid()`

log_normal = tfb.Exp()(tfd.Normal(0, 1))
# ==> `tfd.TransformedDistribution(tfd.Normal(0, 1), tfb.Exp())`

tfb.Exp()([-1., 0., 1.])
# ==> tf.exp([-1., 0., 1.])

__eq__

Return self==value.