treo / rl4j-snake

An *unfinished* RL4J snake playing example

Snake reinforcement learning

This repository contains a simple snake game, and a setup to for an RL agent to learn how to play it.

This uses RL4J and a hilariously bad swing UI to show the game state. Do not consider anything you see here to be good practice.

Note:

• The agents aren't properly tuned
• The A3C algorithm crashes the JVM

Use this at your own risk.