fizzyagent/cs4246-blackjack

RL Black Jack agents trained via DQN and Policy Gradient

Black Jack Reinforcement Learning Agents

This repository contains a set of reinforcement learning agents trained to play the game of Black Jack. The agents are trained using policy gradient and Deep Q-Network.

Setup

Install requirements:

pip install -r requirements.txt

Project Structure

The project is structured as follows:

.
├── game  # contains the Black Jack game environment
├── models  # contains the trained models and training logs
├── training  # contains notebooks used for training and evaluation

Agents

All trained models are located in the models directory. The following models are available:

Agent	Description
dqn_final	Deep Q-Network agent
dqn_final_no_counting	DQN agent trained without card counting features
policy_steps	Policy Gradient agent
policy_steps_no_counting	Policy Gradient agent trained without card counting features

Training logs are also available in the models directory, named as *_training_scores.txt