gfn.distributions

Module Contents

Classes

EmpiricalTerminatingStatesDistribution	Represents an empirical distribution over terminating states.
EmpiricalTrajectoryDistribution	Represents an empirical distribution over trajectories.
TerminatingStatesDistribution	Represents an abstract distribution over terminating states.
TrajectoryBasedTerminatingStateDistribution	Represents a distribution over final states.
TrajectoryDistribution	Represents an abstract distribution over trajectories.

Attributes

TensorPmf

class gfn.distributions.EmpiricalTerminatingStatesDistribution(env, states)

Bases: TerminatingStatesDistribution

Represents an empirical distribution over terminating states.

Parameters

• env (gfn.envs.env.Env) –

• states (gfn.containers.states.States) –

pmf()

Compute the probability mass function of the distribution.

Return type

TensorPmf

class gfn.distributions.EmpiricalTrajectoryDistribution(trajectories)

Bases: TrajectoryDistribution

Represents an empirical distribution over trajectories.

Parameters

trajectories (gfn.containers.trajectories.Trajectories) –

sample(n_trajectories=None)

Sample a batch of trajectories.

Parameters

n_trajectories (Optional[int]) –

Return type

gfn.containers.trajectories.Trajectories

gfn.distributions.TensorPmf

class gfn.distributions.TerminatingStatesDistribution

Bases: abc.ABC

Represents an abstract distribution over terminating states.

abstract pmf()

Compute the probability mass function of the distribution.

Return type

TensorPmf

class gfn.distributions.TrajectoryBasedTerminatingStateDistribution(trajectory_distribution)

Bases: TerminatingStatesDistribution

Represents a distribution over final states.

Parameters

trajectory_distribution (EmpiricalTrajectoryDistribution) –

pmf()

Compute the probability mass function of the distribution.

Return type

TensorPmf

sample(n_final_states=None)

Sample a batch of final states.

Parameters

n_final_states (Optional[int]) –

Return type

gfn.containers.states.States

class gfn.distributions.TrajectoryDistribution

Bases: abc.ABC

Represents an abstract distribution over trajectories.

abstract sample(n_trajectories)

Sample a batch of trajectories.

Parameters

n_trajectories (int) –

Return type

gfn.containers.trajectories.Trajectories