Easy21

Discussion

A complete description of the math: https://github.com/XingyuHe/Easy21/blob/master/ProjectMath.pdf

Usage

Install required packages by running

pip install -r requirements.txt

To use Markov Decision Process and Monte Carlo to estimate state action value function, run

python Markov_Decision_Process.py

To use Monte Carlo Control to estimate state action value function, run

python Monte_Carlo_Control.py

Results

Algorithm	State-Action Value Function	Policy (yellow - hit, purple - stay)
Markov Decision Process
Monte Carlo Control (Linearly Decreasing Epsilon)

Evolution of Monte Carlo Control

Algorithm	State-Action Value Function	Policy (yellow - hit, purple - stay)
Epsilon = 0
Epsilon = 1
Linearly Decreasing Epsilon