Mastering Atari Games with Deep Reinforcement Learning

This work was done as a part of the course project of CS 419 - Introduction to Machine Learning, under Prof. Abir De.

The details about the project, previous work, our contributions, and our results can be found in the report and the presentation

Results

Pong	Breakout	Boxing

Instructions to Run Code

conda create -n env # create new conda environment
conda activate env # activate the conda environment
pip install -r requirements.txt # install dependencies
jupyter notebook <path/to/notebook> # opens the required notebook in jupyter
# run all cells in the notebook

Repository Structure:

.
├── DQN
│   ├── DeepRL_boxing.ipynb
│   ├── DeepRL_breakout.ipynb
│   ├── DeepRL_pong.ipynb
│   ├── Param.py
│   ├── model.py
│   ├── models
│   │   ├── # saved model weights checkpoints for multiple games
│   ├── replay_memory.py
│   ├── utils.py
│   ├── videos
│   │   ├── # videos of agent playing multiple games
│   └── wrappers.py
├── Double-DQN
│   ├── DeepRL_boxing_modified.ipynb
│   ├── DeepRL_breakout_modified.ipynb
│   ├── DeepRL_pong_modified.ipynb
│   ├── Param_modif.py
│   ├── model.py
│   ├── models_modified
│   │   ├── # saved model weights checkpoints for multiple games
│   ├── replay_memory.py
│   ├── utils_modif.py
│   ├── videos_modified
│   │   ├── # videos of agent playing multiple games
│   └── wrappers.py
├── README.md
├── report.pdf
└── requirements.txt

Note: The two DQN and Double-DQN folders in the top directory are standalone folders containing the respective implementation. Also, each game has a separate IPython notebook as indicated in the tree above.