This folder contains the implementation of the Easy21 assignment of David Silver's Reinforcement Learning course.
The script makes use of the following dependencies, which are to be installed by the user using the standard pip module. These dependencies were tested under Ubuntu 18.04 and using Python3.
-
Create a virtual environment:
$ virtualenv pyeasy21
-
Activate the virtual environment:
$ source pyeasy21/bin/activate
-
Install all the dependencies necessary for running RL algorithms and plot results:
$ pip install -r requirements.txt $ pip install -e gym-easy21
The gym-easy21
environment is OpenAI Gym compatible. You can create an instance of it with:
python gym.make('gym_easy21:easy21-v0')
The following scripts implements the 3 algorithms required by the assignment:
mc.py
: Monte Carlo controlsarsa_lambda.py
: TD Learninglinear_approx.py
: Linear Function Approximation