This repository contains the solution to Easy 21 assignment from David Silver's Reinforcement Learning Series. The actual assignment can be found here.
Algorithms implemented include
- Monte Carlo Evaluation
- Monte Carlo Control
- TD Control (SARSA
$\lambda$ with eligibility traces) - Linear Function Approximation
All of these are implemented in Easy 21, a spin-off of the classic Blackjack, with non-standard rules.
requirements.txt contains all the dependencies, and to install them run
pip install -r requirements.txt
in the root directory of the project.
To reproduce the results, try running,
python code/[agent].py
where agents include the ones defined above.