Repository implementing deep reinforcement learning and conducting experiments on the OpenAI environments

# Clone repo
git clone [email protected]:eugeneyan/deep_rl.git && cd deep_rl

# Create conda environment
conda env create -f=environment.yml

# Activate environment
source activate deep_rl

If a box2D error is encountered (e.g., AttributeError: module '_Box2D' has no attribute 'RAND_LIMIT_swigconstant'), please follow the steps below:

pip uninstall Box2D box2d-py
git clone https://github.com/pybox2d/pybox2d
cd pybox2d/
python setup.py clean
python setup.py build
python setup.py install

More details here: openai/gym#100

Currently, the following environments are supported:

• CartPole
• LunarLander

• Deep Q-Network (vanilla) dqn_plain.py
• Deep Q-Network (with target network) dqn.py
• Double Deep Q-Network ddqn.py

Running experiments is a simple as

# To run experiment on cartpole environment
python cartpole_runner.py

# To run experiment on lunar lander environment 
python lunarlander_runner.py

python lunarlander_runner.py --help

usage: lunarlander_runner.py [-h] [--model MODEL] [--render-env RENDER_ENV]
                             [--render-freq RENDER_FREQ]

optional arguments:
  -h, --help            show this help message and exit
  --model MODEL         DeepRL model to use (options: dqn_plain, dqn, ddqn;
                        default: ddqn)
  --render-env RENDER_ENV
                        Whether to render the environment (default: y)
  --render-freq RENDER_FREQ
                        How frequently to render the env (default: 500)
                        --render-env must be set to "y" to render environment

• J. Gao, R. Evans. “DeepMind AI Reduces Google Data Centre Cooling Bill by 40%”, https://deepmind.com/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-40/, 18 Mar 2018

• J. Gao, “Machine Learning Applications for Data Center Optimization”, http://static.googleusercontent.com/media/www.google.com/en//about/datacenters/efficiency/internal/assets/machine-learning-applicationsfor-datacenter-optimization-finalv2.pdf, 18 Mar 2018

• J. Zhao, “Deep Reinforcement Learning for Sponsored Search Real-time Bidding”, arXiv:1803.00259, 2018

• C. Watkins “Learning from delayed rewards”, PhD Thesis, University of Cambridge, England, 1989

• V. Mnih, et al. “Human-level control through deep reinforcement learning.” Nature 518(7540):529-533, 2015.

• V. Minh, et al. “Playing atari with deep reinforcement learning.” CoRR abs/1312.5602, 2013.

• H. van Hasselt, A. Guez, and D. Silver. “Deep reinforcement learning with double Q-Learning.” CoRR abs/1509.06461, 2015