tf_agents.utils.common.entropy

Computes total entropy of distribution.

tf_agents.utils.common.entropy(
    distributions, action_spec, outer_rank=None
)

Args
distributions	A possibly batched tuple of distributions.
action_spec	A nested tuple representing the action spec.
outer_rank	Optional outer rank of the distributions. If not provided use distribution.mode() to compute it.

Returns
A Tensor representing the entropy of each distribution in the batch. Assumes actions are independent, so that marginal entropies of each action may be summed.