minor

diff --git a/src/qtable.py b/src/qtable.py
index 5bb1c519f29001855a9b6e2618b6cd7c04663f84..fb650085955820d25d7fa0194763f9a384e79807 100755 (executable)
--- a/src/qtable.py
+++ b/src/qtable.py
@@ -6,7 +6,7 @@ import sys
 
 from game import Game
 
-learning_rate = 1.0
+learning_rate = 0.001
 discount_factor = 1.0
 
 states_dim = 36864 # 2^10 * 6^2
@@ -56,14 +56,14 @@ for i in range(num_episodes):
             old_score = g.get_score()
             g.shut(opt)
             g.dice()
-            reward = (g.get_score() - old_score) / 11.0
+            reward = g.get_score() - old_score
             new_state_qid = find_state_qid(g.get_shutable(), g.get_diced())
             Q[state_qid, opt_qid] += \
                 learning_rate * (reward
                                  + discount_factor * np.max(Q[new_state_qid, :])
                                  - Q[state_qid, opt_qid])
             state_qid = new_state_qid
-    Q[state_qid, :] = 0
+    Q[state_qid, :] = 0.0
     running_score[0] *= 0.99999999
     running_score[0] += g.get_score()
     running_score[1] *= 0.99999999