WE

Code for ICLR2025 Paper: Rethinking Neural Multi-Objective Combinatorial Optimization via Neat Weight Embedding

Quick Start For WE-Add or WE-CA

• To train a model, such as MOTSP with 20 nodes, run train_motsp_n20.py in the corresponding folder.
• To test a model, such as MOTSP with 20 nodes, run test_motsp_n20.py in the corresponding folder.
• Pretrained models for each problem can be found in the result folder.

Quick Start For WE-CA-U

• To train a unified model, such as MOTSP, run train_motsp.py in the WE-CA/POMO-U folder.
• To test a model, such as MOTSP with 20 nodes, run test_motsp_n20.py in the WE-CA/POMO-U folder.
• Pretrained models for each problem can be found in the WE-CA/POMO-U/result folder.

Reference

If our work is helpful for your research, please cite our paper:

@inproceedings{chen2025rethinking,
  title={Rethinking Neural Multi-Objective Combinatorial Optimization via Neat Weight Embedding},
  author={Chen, Jinbiao and Cao, Zhiguang and Wang, Jiahai and Wu, Yaoxin and Qin, Hanzhang and Zhang, Zizhen and Gong, Yue-Jiao},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025},
}