Finish project 3.

p3_rl/analysis.py

@@ -63,8 +63,7 @@ def question3e():
 def question6():
     answerEpsilon = None
     answerLearningRate = None
-    return answerEpsilon, answerLearningRate
-    # If not possible, return 'NOT POSSIBLE'
+    return 'NOT POSSIBLE'
 
 if __name__ == '__main__':
     print 'Answers to analysis questions:'

p3_rl/qlearningAgents.py

@@ -41,8 +41,7 @@ class QLearningAgent(ReinforcementAgent):
     def __init__(self, **args):
         "You can initialize Q-values here..."
         ReinforcementAgent.__init__(self, **args)
-
-        "*** YOUR CODE HERE ***"
+        self.values = util.Counter()
 
     def getQValue(self, state, action):
         """
@@ -50,9 +49,7 @@ class QLearningAgent(ReinforcementAgent):
           Should return 0.0 if we have never seen a state
           or the Q node value otherwise
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
-
+        return self.values[(state, action)]
 
     def computeValueFromQValues(self, state):
         """
@@ -61,8 +58,10 @@ class QLearningAgent(ReinforcementAgent):
           there are no legal actions, which is the case at the
           terminal state, you should return a value of 0.0.
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        actions = self.getLegalActions(state)
+        if not actions:
+            return 0.0
+        return max([self.getQValue(state, action) for action in actions])
 
     def computeActionFromQValues(self, state):
         """
@@ -70,8 +69,13 @@ class QLearningAgent(ReinforcementAgent):
           are no legal actions, which is the case at the terminal state,
           you should return None.
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        actions = self.getLegalActions(state)
+        if not actions:
+            return None
+        valueActions = [(self.getQValue(state, action), action)
+                        for action in actions]
+        maxValue, maxAction = max(valueActions)
+        return maxAction
 
     def getAction(self, state):
         """
@@ -86,10 +90,12 @@ class QLearningAgent(ReinforcementAgent):
         """
         # Pick Action
         legalActions = self.getLegalActions(state)
-        action = None
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
-
+        if not legalActions:
+            action = None
+        elif util.flipCoin(self.epsilon):
+            action = random.choice(legalActions)
+        else:
+            action = self.computeActionFromQValues(state)
         return action
 
     def update(self, state, action, nextState, reward):
@@ -101,8 +107,10 @@ class QLearningAgent(ReinforcementAgent):
           NOTE: You should never call this function,
           it will be called on your behalf
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        sample = reward + self.discount * self.computeValueFromQValues(nextState)
+        value = self.getQValue(state, action)
+        newValue = (1 - self.alpha) * value + self.alpha * sample
+        self.values[(state, action)] = newValue
 
     def getPolicy(self, state):
         return self.computeActionFromQValues(state)
@@ -164,15 +172,24 @@ class ApproximateQAgent(PacmanQAgent):
           Should return Q(state,action) = w * featureVector
           where * is the dotProduct operator
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        features = self.featExtractor.getFeatures(state, action)
+        value = 0
+        for feature, featureValue in features.items():
+            value += featureValue * self.weights[feature]
+        return value
+
 
     def update(self, state, action, nextState, reward):
         """
            Should update your weights based on transition
         """
-        "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        features = self.featExtractor.getFeatures(state, action)
+        qMax = self.computeValueFromQValues(nextState)
+        difference = (reward + self.discount * qMax) - self.getQValue(state, action)
+        for feature, featureValue in features.items():
+            weight = self.weights[feature]
+            newWeight = weight + self.alpha * difference * featureValue
+            self.weights[feature] = newWeight
 
     def final(self, state):
         "Called at the end of each game."