Learning to Select Cuts for Efficient Mixed-Integer Programming

This is the code for implementing Cut Ranking presented in the paper: Learning to Select Cuts for Efficient Mixed-Integer Programming.

It is based on the open-source MIP solver Python-MIP: python-mip.

Installation

• To install Python-MIP, run:

pip install mip

• Main Python Dependencies: Python (3.6), tensorflow (1.x)

Quick Start

Collect Training Data

To collect training bags and labels, set cut_choice_method to "CollectLabels" (Line 183 of collect_labels.py) and set the problem_idx (Line 174 of collect_labels.py), and run:

python collect_labels.py

Load Training Data

To test loading the data, set the correct training data path (Line 15 of data_loader.py), and run:

python data_loader.py

Train the Cut Ranking Model

To train the model, set the saved model path (Line 66 of train.py), and run:

python train.py

Test on new instances

To test the trained model, you can set the path of trained models in choose_cut.py (Line 312-313), and set cut_choice_method to "Test" (Line 183 of collecet_labels.py), and run:

python collect_labels.py

Paper citation

If you used this code for your experiments or found it helpful, consider citing the following paper:

@misc{huang2021learning,
      title={Learning to Select Cuts for Efficient Mixed-Integer Programming}, 
      author={Zeren Huang and Kerong Wang and Furui Liu and Hui-ling Zhen and Weinan Zhang and Mingxuan Yuan and Jianye Hao and Yong Yu and Jun Wang},
      year={2021},
      eprint={2105.13645},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}