Code accompanying the paper Learning 3-opt heuristics for traveling salesman problem via deep reinforcement learning.
python 3.8.10
Torch
Numpy
tqdm
Apex
pyconcorde
Matplotlib
Train the model using:
python PGTSP20.py
python PGTSP50_100.py
Evaluate the model using:
python TestLearnedAgent.py --load_path <model directory>/<model name>.pt --n_points <number of nodes> --test_size <number of instances in testset> --render
If you this code is useful in your research, please cite our paper:
@inproceedings{sui2021learning,
title={Learning 3-opt heuristics for traveling salesman problem via deep reinforcement learning},
author={Sui, Jingyan and Ding, Shizhe and Liu, Ruizhi and Xu, Liming and Bu, Dongbo},
booktitle={Asian Conference on Machine Learning},
pages={1301--1316},
year={2021},
organization={PMLR}
}
This project extends the foundations established by the previous study "Learning 2-opt Heuristics for the TSP via Deep Reinforcement Learning".