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.
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.
This image showcases an example visual representation of data, illustrating how data can be effectively communicated through imagery.
- 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.
- Install dependencies:
pip install -r requirements.txt