Neurosymbolic Programming Experiments

This repository hosts experiments exploring neurosymbolic programming approaches, with a focus on trip planning applications. The project demonstrates various techniques combining neural networks with symbolic reasoning.

Trip Planning Experiments

The trip-planning folder contains several experimental approaches to neurosymbolic trip planning:

1. Basic Neurosymbolic Approach

• deepseek.py: Combines DeepSeek LLM for natural language understanding with symbolic planning
  • Handles entity extraction, trip planning, and response generation
  • Integrates neural and symbolic components in a structured way
• symbolicplanning.py: Basic symbolic trip planner (experimental/demo only)
  • Simple rule-based system with hard-coded options
  • Demonstrates basic symbolic AI concepts
  • WARNING: Limited capabilities - for educational purposes only

2. Monte Carlo Tree Search (MCTS) Approach

• Implements MCTS algorithm for constrained trip planning
• Features:
  • Core MCTS components: selection, expansion, simulation, backpropagation
  • Budget and time constraints handling
  • Plan evaluation based on activity count and budget utilization
• WARNING: Experimental implementation with hard-coded options

3. Z3 Solver Experiments

• Uses Microsoft's Z3 theorem prover for optimization
• Features:
  • Constraint satisfaction and optimization
  • Discrete options modeling for flights, hotels, and activities
  • Budget and time constraints enforcement
  • Resource utilization maximization
• WARNING: Simplified experimental implementation

4. Prover9 Experiments

• Coming soon

Usage

1. Clone the repository
2. Install dependencies using the provided nsai.yml conda environment file: conda/mamba create --name nsai -f nsai.yml
3. Explore the different approaches in the trip-planning folder

License

This project is licensed under the MIT License - see the LICENSE file for details.

Disclaimer

All implementations in this repository are experimental and for educational purposes only. They are not suitable for production use or real-world trip planning.