agent confidence 버그 픽스

howtokim · Jun 7, 2018 · a6f50ac · a6f50ac
1 parent ef77fd9
commit a6f50ac
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 6 deletions.
diff --git a/agent.py b/agent.py
@@ -76,7 +76,7 @@ def decide_action(self, policy_network, sample, epsilon):
             exploration = False
             probs = policy_network.predict(sample)  # 각 행동에 대한 확률
             action = np.argmax(probs)
-            confidence = 1 + probs[action]
+            confidence = probs[action]
         return action, confidence, exploration
 
     def validate_action(self, action):

diff --git a/agent_custom.py b/agent_custom.py
@@ -76,7 +76,7 @@ def decide_action(self, policy_network, sample, epsilon):
             exploration = False
             probs = policy_network.predict(sample)  # 각 행동에 대한 확률
             action = np.argmax(probs) if np.max(probs) > 0.1 else Agent.ACTION_HOLD
-            confidence = 1 + probs[action]
+            confidence = probs[action]
         return action, confidence, exploration
 
     def validate_action(self, action):
@@ -167,7 +167,3 @@ def act(self, action, confidence):
         else:
             delayed_reward = -1
         return self.immediate_reward, delayed_reward
-
-
-
-# confidence = 1 + probs[action] / (np.sum(probs))