Hybrid Pointer Networks

Sep, 2021

Description

PyTorch implementation of "Hybrid Pointer Networks for Traveling Salesman Problems Optimization"
Ahmed Stohy, Heba-Tullah Abdelhakam, Sayed Ali, Mohammed Elhenawy, Abdallah A Hassan, Mahmoud Masoud ,Sebastien Glaser and Andry Rakotonirainy
PLOSONE : https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0260995

Dependencies

• Python >= 3.7.10
• NumPy
• PyTorch == 1.7.1
• Matplotlib (optional, only for plotting)

Benchmark

Sample instance form TSBLIB benchmark data fnl4461

Guidelines

1. Network Training (Tesla P100-PCIE-16GB)
   For training TSP50 : Run notebook hpn-smallsize-sumagg inside small folder
2. Network Testing
   TSPSmall : Run notebook hpn-smallsize-test.
   TSPLarge : Run notebook hpnlarge-test.
   
3. The Checkpoint file besides each training notebook
   

Acknowledgements

Thanks to qiang-ma/graph-pointer-network for getting me started with the code for the Graph Pointer Network.

This repository contains xbresson/TSP_Transformer adaptations, and I'd like to thank you for making your code available.