Tight Certification of Adversarially Trained Neural Networks via Nonconvex Low-Rank Semidefinite Relaxations

Authors: Hong-Ming Chiu, Richard Y. Zhang

Link to Paper

Citation:

@article{chiu2023tight,
  title={Tight Certification of Adversarially Trained Neural Networks via Nonconvex Low-Rank Semidefinite Relaxations},
  author={Chiu, Hong-Ming and Zhang, Richard Y},
  year={2023}
}

Intorduction

This MATLAB program contains the implementation of neural network verification framework proposed in our paper "ATight Certification of Adversarially Trained Neural Networks via Nonconvex Low-Rank Semidefinite Relaxations".

Requirements

• MATLAB (version R2021a or later).
• MATLAB Parallel Computing Toolbox.
• Artelys Knitro (version 13.1 or later).

Author

Name : Hong-Ming Chiu

Email : hmchiu2 [at] illinois.edu

Website : https://hong-ming.github.io

License

MIT License