DictProbabilityDistribution

class maze.distributions.dict.DictProbabilityDistribution(distribution_dict: Dict[str, ProbabilityDistribution])

Dictionary probability distribution.

The respective functions either return

• the per key distribution properties or

• aggregate the properties across the sub-distributions using a reduce_fun such as mean or sum.

Parameters:

distribution_dict – dictionary holding sub-probability distributions.

deterministic_sample() → Dict[str, torch.Tensor]

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface

entropy(reduce_fun: callable = torch.mean) → torch.Tensor

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface

kl(other: DictProbabilityDistribution, reduce_fun: callable = torch.mean) → torch.Tensor

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface

log_prob(actions: Dict[str, torch.Tensor]) → Dict[str, torch.Tensor]

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface

neg_log_prob(actions: Dict[str, torch.Tensor]) → Dict[str, torch.Tensor]

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface

sample() → Dict[str, torch.Tensor]

(overrides ProbabilityDistribution)

implementation of TorchProbabilityDistribution interface