def is_empty(self):
"""
Returns true if the table is empty (or contains only a default assignment),
false otherwise
:return: true if empty, false otherwise
"""
if len(self._table) == 0:
return True
if len(self._table) > 1:
return False
return list(self._table.keys())[0] == Assignment.create_default(
self._head_vars)
def __init__(self, init_state, system, paused):
from dialogue_system import DialogueSystem
if not isinstance(init_state, DialogueState) or not isinstance(
system, DialogueSystem) or not isinstance(paused, bool):
raise NotImplementedError("UNDEFINED PARAMETERS")
"""
Creates the planning process. Timeout is set to twice the maximum sampling
ime. Then, runs the planner until the horizon has been reached, or the
planner has run out of time. Adds the best action to the dialogue state.
:param init_state: initial dialogue state
"""
self.system = system
self.paused = paused
self.is_terminated = False
settings = system.get_settings()
# setting the timeout for the planning
timeout = Settings.max_sampling_time * 2
# if the speech stream is not finished, only allow fast, reactive responses
if init_state.has_chance_node(settings.user_speech):
timeout = timeout / 5.0
# TODO: ScheduledExecutorService
# service.schedule(() -> isTerminated = true, timeout, TimeUnit.MILLISECONDS);
# step 1: extract the Q-values
eval_actions = self.get_q_values(init_state, settings.horizon)
# step 2: find the action with highest utility
best_action = eval_actions.get_best()[0]
if eval_actions.get_util(best_action) < 0.001:
best_action = Assignment.create_default(
best_action.get_variables())
# step 3: remove the action and utility nodes
init_state.remove_nodes(init_state.get_utility_node_ids())
action_vars = init_state.get_action_node_ids()
init_state.remove_nodes(action_vars)
# step 4: add the selection action to the dialogue state
init_state.add_to_state(best_action.remove_primes())
self.is_terminated = True
def get_best(self):
"""
Returns the most likely assignment of values in the table. If none could be
found, returns an empty assignment.
:return: the assignment with highest probability
"""
if len(self._table) == 0:
self.log.warning("table is empty, cannot extract best value")
raise ValueError()
max_prob = -np.inf
max_assignment = None
for assignment in self._table.keys():
prob = self._table[assignment]
if prob > max_prob:
max_prob = prob
max_assignment = assignment
# TODO: check refactor > there is no case of max_assignment is None
return max_assignment if max_assignment is not None else Assignment.create_default(
self._head_vars)