Reinforcement Learning for Snake

The code in this repository is used to train agents to play snake using techniques from reinforcement learning. Tabular Q-learning and Value Function Approximation are implemented (using a simple Neural Network).

Prerequisites

• pygame
  • python3 -m pip install -U pygame --user

Setup

• Clone the repository: https://github.com/areevesman/reinforcement-learning-for-snake.git
• cd reinforcemt-learning-for-snake
• cd code
• To play snake:
  • python play_snake.py
• To train a snake with Q-learning:
  • python train_q_learning.py
• To train a snake with VFA:
  • python train_deep_q_learning.py
• Training output will be saved to a results folder in the working directory

Authors

• Adam Reevesman
• Evan Liu

Acknowledgments

• Snake with pygame tutorial
• LearnSnake: Teaching an AI to play Snake
• Designing AI: Solving Snake with Evolution