REGLO: Provable Neural Network Repair for Global Robustness Properties

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.

Cite

Text

Fu et al. "REGLO: Provable Neural Network Repair for Global Robustness Properties." AAAI Conference on Artificial Intelligence, 2024. doi:10.1609/AAAI.V38I11.29094

Markdown

[Fu et al. "REGLO: Provable Neural Network Repair for Global Robustness Properties." AAAI Conference on Artificial Intelligence, 2024.](https://mlanthology.org/aaai/2024/fu2024aaai-reglo/) doi:10.1609/AAAI.V38I11.29094

BibTeX

@inproceedings{fu2024aaai-reglo,
  title     = {{REGLO: Provable Neural Network Repair for Global Robustness Properties}},
  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},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2024},
  pages     = {12061-12071},
  doi       = {10.1609/AAAI.V38I11.29094},
  url       = {https://mlanthology.org/aaai/2024/fu2024aaai-reglo/}
}