Skip to content

Learning to improve solution quality using Neural-guided Ant Colony Optimization for Travelling Salesman Problem.

License

Notifications You must be signed in to change notification settings

shoraaa/NeuFACO

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

68 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Neural Focused Ant Colony Optimization

Neural Focused Ant Colony Optimization (NeuFACO) is a non-autoregressive framework for solving the Traveling Salesman Problem (TSP). It combines reinforcement learning with advanced Ant Colony Optimization (ACO) techniques to achieve both efficiency and solution quality. The code is built upon the codebase of GFACS, which itself is developed based on the DeepACO codebase.

Key Features

Reinforcement Learning Integration:

Uses Proximal Policy Optimization (PPO) to train a Graph Neural Network (GNN).

Generates instance-specific heuristic guidance rather than relying on fixed training schemes.

Untitled

Enhanced ACO Framework:

Incorporates candidate lists, focused tour modification, and scalable local search.

Efficiently integrates the learned heuristic into the solution process.

Untitled

Performance:

Leverages amortized inference together with the parallel stochastic exploration of ACO.

Untitled

Provides fast and high-quality solutions across diverse TSP instances.

Dependencies

  • Python 3.11.5
  • PyTorch 2.1.1
  • PyTorch Geometric 2.4.0

We strongly recommend using uv for virtual environment in this project.

uv venv python=3.11
uv pip install torch torch_geometric
uv pip install numpy numba pandas pyvrp scipy tdqm wandb

For the complete list of dependencies, please refer to the requirements.txt file.

Usage

For the usage of the code, please refer to each folder's README.md file.

About

Learning to improve solution quality using Neural-guided Ant Colony Optimization for Travelling Salesman Problem.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •