tile-match-gym 0.0.1

A set of reinforcement learning environments for tile matching games, consistent with the OpenAI Gym API.

Tile Matching Reinforcement Learning Environments

Welcome to the Reinforcement Learning Environments for Tile Matching Games repository! This repository provides a collection of tile matching game environments (like Bejeweled or Candy Crush) implemented in NumPy, poised to push reinforcement learning research forwards.

This genre of games is characterised by the following features which we find useful for reinforcement learning research:

• Large action spaces
• Intuitive action hierarchies
• Procedurally generated levels
• Structured complex stochasticity in transition dynamics

Work in Progress - Pre-release

Please note that this project is a work in progress, and while many exciting features are on the roadmap, they might not all be fully implemented at this time.

Installation

On release, the environments will be installable via pip:

pip install tile-match-gym

Citation

If you use this repository please cite as below:

@software{tile_match_gym,
  author = {Patel, Akshil and Elson, James},
  title = {{Tile Matching Game Reinforcement Learning Environments}},
  url = {https://github.com/akshilpatel/tile-match-gym},
  version = {0.0.1},
  year = {2023}
}