fix bug

MorvanZhou · MorvanZhou · commit cfa37b53002f · 2017-03-21T12:23:04.000+11:00
diff --git a/Reinforcement_learning_TUT/5.2_Prioritized_Replay_DQN/RL_brain.py b/Reinforcement_learning_TUT/5.2_Prioritized_Replay_DQN/RL_brain.py
@@ -234,7 +234,7 @@ def build_layers(s, c_names, n_l1, w_initializer, b_initializer):
     def store_transition(self, s, a, r, s_):
         if self.prioritized:    # prioritized replay
             transition = np.hstack((s, [a, r], s_))
-            self.memory.store(0.9, transition)    # have 1 priority for newly arrived transition
+            self.memory.store(1, transition)    # have 1 priority for newly arrived transition
         else:       # random replay
             if not hasattr(self, 'memory_counter'):
                 self.memory_counter = 0
diff --git a/Reinforcement_learning_TUT/5.2_Prioritized_Replay_DQN/run_MountainCar.py b/Reinforcement_learning_TUT/5.2_Prioritized_Replay_DQN/run_MountainCar.py
@@ -68,8 +68,9 @@ def train(RL):
 his_natural = train(RL_natural)
 his_prio = train(RL_prio)
 
-plt.plot(his_natural[0, :], his_natural[1, :], c='b', label='natural DQN')
-plt.plot(his_prio[0, :], his_prio[1, :], c='r', label='DQN with prioritized replay')
+# compare based on first success
+plt.plot(his_natural[0, :], his_natural[1, :] - his_natural[1, 0], c='b', label='natural DQN')
+plt.plot(his_prio[0, :], his_prio[1, :] - his_prio[1, 0], c='r', label='DQN with prioritized replay')
 plt.legend(loc='best')
 plt.ylabel('total training time')
 plt.xlabel('episode')
