def exp3(self, numActions, gamma, rewardMin=0, rewardMax=1):
#weights = [1.0] * numActions
t = 0
while True:
probabilityDistribution = distr(self.weights, gamma)
choice = draw(probabilityDistribution)
self.choiceFreq[choice] += 1
#put choice in the queue
self.choice_queue.put(choice)
self.choice_queue.join()
#get reward from queue
theReward = self.reward_queue.get()
self.reward_queue.task_done()
#theReward = reward(choice, t)
scaledReward = (theReward - rewardMin) / (
rewardMax - rewardMin) # rewards scaled to 0,1
probChoice = float(self.choiceFreq[choice] + 1) / (t + 1)
#estimatedReward = 1.0 * scaledReward / probabilityDistribution[choice]
#estimatedReward = (1.0 * scaledReward) / probChoice
estimatedReward = 1.0 * scaledReward * self.distanceLastchoice(
t, choice)
self.weights[choice] *= math.exp(
estimatedReward * gamma /
numActions) # important that we use estimated reward here!
yield choice, theReward, estimatedReward, self.weights
t = t + 1
def exp3(self,numActions, gamma, rewardMin = 0, rewardMax = 1):
t = 0
while True:
probabilityDistribution = distr(self.weights, gamma)
choice = draw(probabilityDistribution)
self.choiceFreq[choice]+=1
#put choice in the queue
self.choice_queue.put(choice)
self.choice_queue.join()
#get reward from queue
theReward = self.reward_queue.get()
self.reward_queue.task_done()
#theReward = reward(choice, t)
scaledReward = (theReward - rewardMin) / (rewardMax - rewardMin) # rewards scaled to 0,1
#-case frequency
#probChoice=float(self.choiceFreq[choice] +1)/(t+1)
#estimatedReward = (1.0 * scaledReward) / probChoice
#-case distribution
#estimatedReward = 1.0 * scaledReward / probabilityDistribution[choice]
#-case distance
estimatedReward = 1.0 * scaledReward * self.distanceLastchoice(t,choice)
self.weights[choice] *= math.exp(estimatedReward * gamma / numActions) # important that we use estimated reward here!
yield choice, theReward, estimatedReward, self.weights
t = t + 1
def exp3(numActions, reward, gamma, rewardMin = 0, rewardMax = 1):
weights = [1.0] * numActions
t = 0
while True:
probabilityDistribution = distr(weights, gamma)
choice = draw(probabilityDistribution)
theReward = reward(choice, t)
scaledReward = (theReward - rewardMin) / (rewardMax - rewardMin) # rewards scaled to 0,1
estimatedReward = 1.0 * scaledReward / probabilityDistribution[choice]
weights[choice] *= math.exp(estimatedReward * gamma / numActions) # important that we use estimated reward here!
yield choice, theReward, estimatedReward, weights
t = t + 1
def exp3(numActions, reward, gamma, rewardMin=0, rewardMax=1):
weights = [1.0] * numActions
t = 0
while True:
probabilityDistribution = distr(weights, gamma)
choice = draw(probabilityDistribution)
theReward = reward(choice, t)
scaledReward = (theReward - rewardMin) / (rewardMax - rewardMin
) # rewards scaled to 0,1
estimatedReward = 1.0 * scaledReward / probabilityDistribution[choice]
weights[choice] *= math.exp(
estimatedReward * gamma /
numActions) # important that we use estimated reward here!
yield choice, theReward, estimatedReward, weights
t = t + 1
def bidding(self):
bid = draw(self.pi)
return (bid * self.eps)