Merge pull request p-christ#20 from simonalford42/random_episode

fix bug where first episode using Epsilon_Greedy_Exploration is random

Author: p-christ

exploration_strategies/Epsilon_Greedy_Exploration.py

@@ -33,7 +33,7 @@ def perturb_action_for_exploration_purposes(self, action_info):
         epsilon = self.get_updated_epsilon_exploration(action_info)
 
 
-        if (random.random() > epsilon or turn_off_exploration) and (episode_number > self.random_episodes_to_run):
+        if (random.random() > epsilon or turn_off_exploration) and (episode_number >= self.random_episodes_to_run):
             return torch.argmax(action_values).item()
         return random.randint(0, action_values.shape[1] - 1)
 
@@ -65,4 +65,4 @@ def add_exploration_rewards(self, reward_info):
 
     def reset(self):
         """Resets the noise process"""
-        pass
+        pass