MazeBase: A Sandbox for Learning from Games

Sainbayar Sukhbaatar, Arthur Szlam, Gabriel Synnaeve, Soumith Chintala, Rob Fergus
11/23/2015
cs.LGcs.AIcs.NE

Official code: facebookarchive/MazeBase (244 stars, last commit 2019-10-21).

Abstract

This paper introduces MazeBase: an environment for simple 2D games, designed as a sandbox for machine learning approaches to reasoning and planning. Within it, we create 10 simple games embodying a range of algorithmic tasks (e.g. if-then statements or set negation). A variety of neural models (fully connected, convolutional network, memory network) are deployed via reinforcement learning on these games, with and without a procedurally generated curriculum. Despite the tasks' simplicity, the performance of the models is far from optimal, suggesting directions for future development. We also demonstrate the versatility of MazeBase by using it to emulate small combat scenarios from StarCraft. Models trained on the MazeBase version can be directly applied to StarCraft, where they consistently beat the in-game AI.

View on arXivView PDF

Code Implementations(1)

facebookarchive/MazeBaseOfficial100%

Simple environment for creating very simple 2D games and training neural network models to perform tasks within them

24450Lua, PythonDec 17, 20157 years agoNOASSERTION

Similar Papers

StarCraft II: A New Challenge for Reinforcement Learning

89% match1 repo8.3k2017

Nemobot Games: Crafting Strategic AI Gaming Agents for Interactive Learning with Large Language Models

87% match2026

Ludax: A GPU-Accelerated Domain Specific Language for Board Games

87% match2025

Reinforcement Learning in Strategy-Based and Atari Games: A Review of Google DeepMinds Innovations

86% match2025

Exploring Parity Challenges in Reinforcement Learning through Curriculum Learning with Noisy Labels

86% match2023

MinAtar: An Atari-Inspired Testbed for Thorough and Reproducible Reinforcement Learning Experiments

86% match1 repo3352019

Cite this paper

@article{sukhbaatar2015mazebase,
  title  = {MazeBase: A Sandbox for Learning from Games},
  author = {Sainbayar Sukhbaatar and Arthur Szlam and Gabriel Synnaeve and Soumith Chintala and Rob Fergus},
  year   = {2015},
  eprint = {1511.07401},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1511.07401}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1511.07401.svg)](https://codeofpaper.com/paper/1511.07401)

Discussion