ActionDistribution

class posggym.agents.utils.action_distributions.ActionDistribution

Abstract base class for action distributions.

abstract sample() → Any

Sample an action from the distribution.

abstract pdf(action: Any) → float

Get the probability density of an action.

Available Action Distributions

class posggym.agents.utils.action_distributions.DiscreteActionDistribution(probs: Dict[Any, float], rng: seeding.RNG | None = None)

Action distribution for discrete actions.

class posggym.agents.utils.action_distributions.MultiDiscreteActionDistribution(probs: List[Dict[Any, float]], rng: seeding.RNG | None = None)

Action distribution for multi-discrete actions.

class posggym.agents.utils.action_distributions.NormalActionDistribution(mean: float | np.ndarray, stddev: float | np.ndarray, rng: np.random.Generator | None = None)

Action distribution for continuous normally distributed actions.

class posggym.agents.utils.action_distributions.DeterministicActionDistribution(action: int | float | ndarray)

Action distribution for deterministic action distribution.

class posggym.agents.utils.action_distributions.ContinousUniformActionDistribution(low: float | np.ndarray, high: float | np.ndarray, rng: np.random.Generator | None = None)

Action distribution for continuous uniformly distributed actions.

class posggym.agents.utils.action_distributions.DiscreteUniformActionDistribution(low: int | np.ndarray, high: int | np.ndarray, rng: np.random.Generator | None = None)

Action distribution for discrete uniformly distributed actions.

Samples uniformly from the set of integers between low and high, inclusive. Works for both scalar and vector actions.