def sp_func(s: S,
qf=qf,
state_action_func=state_action_func,
softmax=softmax,
epsilon=epsilon) -> Mapping[A, float]:
av_dict = {a: qf((s, a)) for a in state_action_func(s)}
return get_softmax_action_probs(av_dict) if softmax else\
get_epsilon_action_probs(av_dict, epsilon)
def edit_state_action_to_epsilon_greedy(
self,
state: S,
action_value_dict: Mapping[A, float],
epsilon: float
) -> None:
self.policy_data[state] = get_epsilon_action_probs(
action_value_dict,
epsilon
)
def get_soft_policy_from_qf_dict(qf_dict: SAf, softmax: bool,
epsilon: float) -> Policy:
if softmax:
ret = Policy(
{s: get_softmax_action_probs(v)
for s, v in qf_dict.items()})
else:
ret = Policy({
s: get_epsilon_action_probs(v, epsilon)
for s, v in qf_dict.items()
})
return ret
def get_epsilon_policy_from_qf(qf_dict: Mapping[S, Mapping[A, float]],
epsilon: float) -> Policy:
return Policy(
{s: get_epsilon_action_probs(v, epsilon)
for s, v in qf_dict.items()})
def sp_func(s: S, softmax=softmax, epsilon=epsilon) -> Mapping[A, float]:
av_dict = get_act_value_dict_from_state(s)
return get_softmax_action_probs(av_dict) if softmax else\
get_epsilon_action_probs(av_dict, epsilon)
