def getAction(self, stateTime, stateDay, stateLocation, stateActivity,
stateLastNotification):
"""
The function feedObservation() receives the 4-tuple elements (i.e., time, location,
activity, and time elapsed since last notification) and makes a decision of sending a
notificatiion or not.
The function is anticipated to be provided implementation
Returns:
A bool indicating whether to send the notification or not
"""
if self.operatingMode == BaseAgent.MODE_ITERATIVE:
# check stage
if self.stage != BaseAgent.STAGE_WAIT_ACTION:
raise Exception("It is not in the stage of determining action")
self.stage = BaseAgent.STAGE_WAIT_REWARD
# check argument value
if stateTime not in utils.allTimeStates():
raise Exception("Invalid stateTime value (got %d)" % stateTime)
if stateDay not in utils.allDayStates():
raise Exception("Invalid stateDay value (got %d)" % stateDay)
if stateLocation not in utils.allLocationStates():
raise Exception("Invalid stateLocation value (got %d)" %
stateLocation)
if stateActivity not in utils.allActivityStates():
raise Exception("Invalid stateActivity value (got %d)" %
stateActivity)
if stateLastNotification not in utils.allLastNotificationStates():
raise Exception("Invalid stateActivity value (got %d)" %
stateLastNotification)
def generateInitialModel(self):
self.qTable = {}
for sTime in utils.allTimeStates():
for sDay in utils.allDayStates():
for sLocation in utils.allLocationStates():
for sActivity in utils.allActivityStates():
for sNotification in utils.allLastNotificationStates():
state = (sTime, sDay, sLocation, sActivity, sNotification)
self.qTable[state] = {a: 0.0 for a in [True, False]}
self.numSteps = 0
def __init__(self):
self.behavior = {}
for sTime in utils.allTimeStates():
for sDay in utils.allDayStates():
for sLocation in utils.allLocationStates():
for sActivity in utils.allActivityStates():
for sNotification in utils.allLastNotificationStates():
state = (sTime, sDay, sLocation, sActivity,
sNotification)
self.behavior[state] = (random.random() < 0.5)
def __init__(self, deviationProb=0.1):
self.behavior = {}
for sTime in utils.allTimeStates():
for sDay in utils.allDayStates():
for sLocation in utils.allLocationStates():
for sActivity in utils.allActivityStates():
for sNotification in utils.allLastNotificationStates():
state = (sTime, sDay, sLocation, sActivity,
sNotification)
self.behavior[state] = (random.random() < 0.5)
self.probTake = 1. - deviationProb
self.probNotTake = deviationProb
def generateInitialModel(self):
self.qTable = {}
for sTime in utils.allTimeStates():
for sDay in utils.allDayStates():
for sLocation in utils.allLocationStates():
for sActivity in utils.allActivityStates():
for sNotification in utils.allLastNotificationStates():
state = (sTime, sDay, sLocation, sActivity,
sNotification)
#self.qTable[state] = {a: np.random.random() * 1e-3 for a in [True, False]}
self.qTable[state] = {True: 1e-5, False: 0.}
self.lastState = None
self.lastAction = None
self.lastReward = None
self.numSteps = 0