tfp.experimental.substrates.numpy.distributions.RegisterKL

View source on GitHub

Class RegisterKL

Decorator to register a KL divergence implementation function.

Usage:

@distributions.RegisterKL(distributions.Normal, distributions.Normal) def _kl_normal_mvn(norm_a, norm_b): # Return KL(norm_a || norm_b)

__init__

View source

__init__(
    dist_cls_a,
    dist_cls_b
)

Initialize the KL registrar.

Args:

  • dist_cls_a: the class of the first argument of the KL divergence.
  • dist_cls_b: the class of the second argument of the KL divergence.

Methods

__call__

View source

__call__(kl_fn)

Perform the KL registration.

Args:

  • kl_fn: The function to use for the KL divergence.

Returns:

kl_fn

Raises:

  • TypeError: if kl_fn is not a callable.
  • ValueError: if a KL divergence function has already been registered for the given argument classes.