Module: tfp.experimental.bijectors

TensorFlow Probability experimental bijectors package.

Classes

class HighwayFlow: Implements an Highway Flow bijector [1].

class ScalarFunctionWithInferredInverse: Bijector to associate a numeric inverse with any invertible function.

class Sharded: A meta-bijector meant for use in an SPMD distributed context.

Functions

build_trainable_highway_flow(...): Builds a HighwayFlow parameterized by trainable variables.

forward_log_det_jacobian_ratio(...): Computes p.fldj(x, ndims) - q.fdlj(y, ndims), numerically stably.

inverse_log_det_jacobian_ratio(...): Computes p.ildj(x, ndims) - q.idlj(y, ndims), numerically stably.

make_distribution_bijector(...): Builds a bijector to approximately transform N(0, 1) into distribution.