vcharraut / dqn-atari

DQN implementation in the Atari environment

DQN-Atari

The goal of this repo is to implement the DQN algorithm and try to benchmarks it in the atari environment made by OpenAI.
Made in Python with the framework PyTorch.

Value based:

• DQN [1]
• Double DQN [2]
• Dueling Network Architecture [3]

---

Setup

Clone the code repo and install the requirements.

git clone https://github.com/VCanete/DRL-Atari.git
cd DRL-Atari
python setup.py install
pip install -r requirements.txt

---

Run the agents

python main.py --do play --env breakout --algo dueling

The recording are saved in playground/$env_name/recording

---

Results

---

Acknowledgements

• @lilianweng for DQN, Double DQN, Dueling Network Architecture
• @fg91 for DQN, Double DQN, Dueling Network Architecture

This project was made under the supervision of Arthur Aubret for the AI master's degree of the University Lyon 1.

References

[1] Playing Atari with Deep Reinforcement Learning
[2] Deep Reinforcement Learning with Double Q-learning
[3] Dueling Network Architectures for Deep Reinforcement Learning