REGLO: Provable Neural Network Repair for Global Robustness Properties
Feisi Fu, Zhilu Wang, Weichao Zhou, Yixuan Wang, Jiameng Fan, Chao Huang, Qi Zhu, Xin Chen
Abstract
We present REGLO, a novel methodology for repairing pretrained neural networks to satisfy global robustness and individual fairness properties. A neural network is said to be globally robust with respect to a given input region if and only if all the input points in the region are locally robust. This notion of global robustness also captures the notion of individual fairness as a special case. We prove that any counterexample to a global robustness property must exhibit a corresponding large gradient. For ReLU networks, this result allows us to efficiently identify the linear regions that violate a given global robustness property. By formulating and solving a suitable robust convex optimization problem, REGLO then computes a minimal weight change that will provably repair these violating linear regions.
BibTeX
@article{Fu_Wang_Zhou_Wang_Fan_Huang_Zhu_Chen_Li_2024, title={REGLO: Provable Neural Network Repair for Global Robustness Properties}, volume={38}, url={https://ojs.aaai.org/index.php/AAAI/article/view/29094}, DOI={10.1609/aaai.v38i11.29094}, abstractNote={We present REGLO, a novel methodology for repairing pretrained neural networks to satisfy global robustness and individual fairness properties. A neural network is said to be globally robust with respect to a given input region if and only if all the input points in the region are locally robust. This notion of global robustness also captures the notion of individual fairness as a special case. We prove that any counterexample to a global robustness property must exhibit a corresponding large gradient. For ReLU networks, this result allows us to efficiently identify the linear regions that violate a given global robustness property. By formulating and solving a suitable robust convex optimization problem, REGLO then computes a minimal weight change that will provably repair these violating linear regions.}, number={11}, journal={Proceedings of the AAAI Conference on Artificial Intelligence}, author={Fu, Feisi and Wang, Zhilu and Zhou, Weichao and Wang, Yixuan and Fan, Jiameng and Huang, Chao and Zhu, Qi and Chen, Xin and Li, Wenchao}, year={2024}, month={Mar.}, pages={12061-12071} }