omss 0.1.0

Python package for generating customizable matrix reasoning puzzles, inspired by Raven Progressive Matrices

omss is a Python package for generating matrix reasoning puzzles, inspired by Raven's Progressive Matrices. It allows users to generate an unlimited number of customizable puzzles across a range of difficulty levels by setting rules for visual elements. Please check out the Documentation for more information.

Features

• Customizable matrix reasoning puzzle generation
• Reproducibility with seed control
• Colorblind-friendly visual design
• 5 different rules: distribute_three, progression, arithmetic, constant, full_constant
• Generate virtually unlimited unique puzzle variations
• Includes ~80 predefined rulesets across 6 difficulty levels, each of which can produce a huge variety of distinct puzzles

             

Installation

pip install omss

Quick start

#import statements
import omss
from omss import Ruletype, AttributeType, Rule, create_matrix, plot_matrices, ruleset

#define the rules for the puzzle
rules = {
    'BigShape': [       
        Rule(Ruletype.DISTRIBUTE_THREE, AttributeType.SHAPE),
        Rule(Ruletype.CONSTANT, AttributeType.ANGLE),
        Rule(Ruletype.CONSTANT, AttributeType.COLOR),
        Rule(Ruletype.CONSTANT, AttributeType.NUMBER),
        Rule(Ruletype.FULL_CONSTANT, AttributeType.SIZE, value = 'medium')]}
    
#create the matrices and alternatives
solution_matrix, problem_matrix, alternatives = create_matrix(rules, alternatives =4, save = False)

#plot the matrices and alternatives
plot_matrices(solution_matrix, problem_matrix, alternatives)

                 

       

Documentation

For full examples and advanced usage, see the full tutorial and documentation: Tutorial and documentation

License

This project is licensed under the terms of the GNU license: LICENSE.

Acknowledgements

This project was funded by the NWO Open Science grant (OSF23.2.029: Open Matrices: A global, free resource for testing cognitive ability) and the Netherlands eScience Center fellowship of Nicholas Judd.

The package itself was inspired in part by raven-gen. Chi Zhang, Feng Gao, Baoxiong Jia, Yixin Zhu, Song-Chun Zhu Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019

Aran van Hout, Jordy van Langen, Rogier Kievit, Nicholas Judd