Deep-Reinforcement-Learning
-
Deep-Reinforcement-Learning
-
MP5.ipynb
-
README.md
-
agent.py
-
agent_double.py
-
config.py
-
dqn.gif
-
memory.py
-
model.py
-
utils.py
README.md
Deep-Reinforcement-Learning
- Implemented Deep Q-Network (DQN) and Double DQN on the game of Breakout using the OpenAI Gym.
- Double DQN uses a target network to calculate the target Q-value (next state maximum) whereas the vanilla DQN uses the same network for computing Q values for both the current and next state.
- All calculations were run on Google Colab. Due to limitations in availability of computational resources, the model could be trained only for 500 episodes.