def _get_abstract_state(self, abstract_instance):
if self.template_policy.get_process_likelihood(abstract_instance) >= self.min_likelihood:
self.template_policy.process_integrate(abstract_instance)
return self.template_policy
candidates, likelihood = multi_max(self.parent.all_states, key=abstract_instance.get_process_likelihood)
if likelihood >= self.min_likelihood:
abstract_state = random.choice(candidates)
abstract_state.process_integrate(abstract_instance)
return abstract_state
return abstract_instance
def _get_next_motor(self, state):
candidates, _ = multi_max([(state, a) for a in self.actions], key=self.template_policy.get_value)
_, motor = random.choice(candidates)
return motor
def _get_next_state(self, pair, sub_pair, sub_state):
key_func = lambda x: self.get_total_likelihood(pair, x) * x.get_transition_likelihood(sub_pair, sub_state)
candidates, _ = multi_max(self.all_states, key=key_func)
return random.choice(candidates)