[AAAI 2025] Destroy and Repair Using Hyper-Graphs for Routing

The code can only be used for non-commercial purposes. Please contact the authors if you want to use this code for business matters. If this repository is helpful for your research, please cite our paper:

@article{Li_Liu_Wang_Zhang_2025,   
title={Destroy and Repair Using Hyper-Graphs for Routing},  
volume={39},  
number={17}, 
journal={Proceedings of the AAAI Conference on Artificial Intelligence}, 
author={Li, Ke and Liu, Fei and Wang, Zhenkun and Zhang, Qingfu}, 
year={2025}, 
pages={18341-18349}}

dependencies

• python>=3.8
• Pytorch 1.12.1 or 1.13
• numpy==1.23.3
• matplotlib==3.5.2
• tqdm==4.64.1
• pytz==2022.1

How to use

Resources

• training data
  The same as that in LEHD.
  you can download the data from
  https://drive.google.com/drive/folders/1LptBUGVxQlCZeWVxmCzUOf9WPlsqOROR?usp=sharing
  or
  https://pan.baidu.com/s/12uxjol_5pAlnm0j4F6D_RQ?pwd=rzja

• finetuning data
  https://drive.google.com/file/d/1aU1Kpqfy2bMgY1lYdQzjdBN87W3dzE9M/view?usp=sharing

• testing data
  in ./TSP/data or ./CVRP/data

Testing

cd TSP
# for TSP100 etc
python test.py 
# for TSPlib
python test_tsplib.py 

Training

cd TSP
python train.py

For CVRP, it's similar.

Acknowledgements

DRHG's code implementation is based on the code of POMO and LEHD. Thanks to them.