visual-ds / deep-reinforcement-learning

Scientific Initiation in Deep Reinforcement Learning (2019 - 2020, FGV-EMAp)

deep-reinforcement-learning

Scientific Initiation (2019 - 2020) in Deep Reinforcement Learning at Getulio Vargas Foundation, supervised by Dr. Jorge Poco.

If you wanna know the current state of my studies, see my presentation.

In this repository, you'll find implementations of:

• CartPole-v0 random search policy
• FrozenLake8x8-v0 value iteration
• FrozenLake8x8-v0 policy iteration
• MountainCar-v0 Q-learning
• MountainCar-v0 deep Q-learning

Installation

Clone this repository:

git clone https://github.com/lucasresck/deep-reinforcement-learning.git

Usage

All Python files can be run in this way:

python mountaincar_deep_q_learning.py

Some files have special arguments:

MountainCar-v0 deep Q-learning

For training and saving the model:

python mountaincar_deep_q_learning.py

For running the saved model:

python mountaincar_deep_q_learning.py --run

For saving a video from the trained model:

python mountaincar_deep_q_learning.py --run --record

MountainCar-v0 Q-learning

For training and saving the model:

python mountaincar_q_learning.py

For running the saved model:

python mountaincar_q_learning.py --run