def mdp_rep_to_mrp_rep1(
mdp_rep: SASf,
policy_rep: SAf
) -> SSf:
return {s: sum_dicts([{s1: p * v2 for s1, v2 in v[a].items()}
for a, p in policy_rep[s].items()])
for s, v in mdp_rep.items()}
def get_state_reward_gen_dict(tr: SASf, rr: SASf) \
-> Mapping[S, Mapping[A, Callable[[], Tuple[S, float]]]]:
return {
s: {
a: get_state_reward_gen_func(tr[s][a], rr[s][a])
for a, _ in v.items()
}
for s, v in rr.items()
}
def flatten_sasf_dict(sasf: SASf) -> FlattenedDict:
return [((s, a, s1), f)
for s, asf in sasf.items()
for a, sf in asf.items()
for s1, f in sf.items()]