Scalable Kernel Inverse Optimization

Official implementation of the NeurIPS 2024 paper Scalable Kernel Inverse Optimization.

Installation

pip install -r requirements.txt

Content

• algorithm.py This is the main executable python file. The dataset used for training and the relevant hyperparameters are specified in this file. Subsequently, the SCS/SSO solver is invoked to solve the inverse optimization problem.

• For example, you can run the code with python algorithm.py --k 1e-6 --scaler_d_T 100 --score 0 --env 6 --it 20 --batch 10000

• cvx_solver.py Construct the inverse optimization problem using CVXPY and solve it with the SCS solver.

• cd_solver.py Use SSO algorithm to solve the inverse optimization problem in a distributed fashion. At every iteration, use CVXPY to model the sub problem and use SCS to solve it.

Citation

@article{long2024scalable,
  title={Scalable kernel inverse optimization},
  author={Long, Youyuan and Ok, Tolga and Zattoni Scroccaro, Pedro and Mohajerin Esfahani, Peyman M},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={99464--99487},
  year={2024}
}