Merge pull request #14 from quanvuong/master

Consistent initial type (float) for episode_rewards

baselines/deepq/simple.py

@@ -222,7 +222,7 @@ def learn(env,
             episode_rewards[-1] += rew
             if done:
                 obs = env.reset()
-                episode_rewards.append(0)
+                episode_rewards.append(0.0)
 
             if t > learning_starts and t % train_freq == 0:
                 # Minimize the error in Bellman's equation on a batch sampled from replay buffer.