Skip to content

NullLabTests/AlphaGeometryPlusPlus

Repository files navigation

AlphaGeometry++

AlphaGeometry++: A Hybrid Neuro-Symbolic Geometry Solver with Enhanced Adaptability and Explainability

WARNING THIS IS A WORK IN PROGRESS: I will return to this when I can.

Overview

AlphaGeometry++ builds upon DeepMind's AlphaGeometry by integrating a neural language model with a symbolic deduction engine and a Warren Abstract Machine (WAM) for efficient logical inference. The project aims to solve complex geometry problems with improved adaptability and explainability.



Example Visual Representations of Data
This image showcases an example visual representation of data, illustrating how data can be effectively communicated through imagery.


Key Features

  • Neural Language Model: Suggests geometric constructions.
  • Symbolic Deduction Engine (DDAR): Provides rigorous geometric proofs.
  • WAM Integration: Executes efficient logical inferences.
  • Neuro-Symbolic Bridge: Combines LM suggestions with symbolic reasoning.
  • Meta-Learning: Adapts the solver to new problem types.
  • Explainability: Outputs human-readable proofs and visualizations.

Setup & Usage

  1. Install dependencies:
    pip install -r requirements.txt

About

AlphaGeometry++: A Hybrid Neuro-Symbolic Geometry Solver with Enhanced Adaptability and Explainability

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages