Regularizers that penalize big-M constants, unstable neurons, and per-sample LP relaxation gaps during neural network training reduce MILP solve times by up to four orders of magnitude while preserving surrogate accuracy.
Cer- tified robustness to data poisoning in gradient-based training.arXiv preprint arXiv:2406.05670
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
representative citing papers
A barrier-certificate framework certifies non-trivial robustness radii for neural networks under worst-case l_p poisoning during training and at test time, with PAC bounds derived via scenario convex programming.
citing papers explorer
-
Relaxation-Informed Training of Neural Network Surrogate Models
Regularizers that penalize big-M constants, unstable neurons, and per-sample LP relaxation gaps during neural network training reduce MILP solve times by up to four orders of magnitude while preserving surrogate accuracy.
-
Robustness Certificates for Neural Networks against Adversarial Attacks
A barrier-certificate framework certifies non-trivial robustness radii for neural networks under worst-case l_p poisoning during training and at test time, with PAC bounds derived via scenario convex programming.