Robust CurveMoE: Multi-Norm Adversarial Defense for Mixture-of-Experts Models via Mode Connectivity
2026-08-26 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address the challenge of making neural networks resistant to different kinds of small changes (perturbations) that might confuse them. Instead of trying to handle all types of changes with one model, they create a system called Robust CurveMoE, which links specialized models for each change type along a smooth path. This approach shares parts of the model between them to save resources and improves defense against attacks while keeping accuracy high. They also introduce smart methods to update only important parts during training, making the process faster and more efficient. Tests on popular datasets show their method works better than previous techniques.
Multi-norm adversarial defenseNeural networksMixture-of-expertsPerturbation normsCurve optimizationSelective expertizationPartial updatingWideResNetVision TransformerCIFAR-100
Authors
Xu Zhang, Ren Wang
Abstract
Multi-norm adversarial defense aims to protect neural networks against perturbations defined by different norm constraints, but existing methods typically optimize competing robustness objectives within a single parameter configuration, leading to substantial training cost and unfavorable robustness trade-offs. We propose Robust CurveMoE, an efficient mixture-of-experts framework that connects models specialized for different perturbation norms through a low-loss path and exploits the complementary robustness profiles of models along this path. Robust CurveMoE derives clean and norm-specialized experts from robustness-constrained curve locations and selectively expertizes only influential layers, while sharing the remaining parameters across routing paths. To further reduce curve-construction cost, we introduce contribution-guided partial updating, which selects influential curve parameters using initialization-based gradient scores. We also theoretically bound the objective gap between partial and full curve optimization. Experiments on CIFAR-100 and ImageNet-100 with WideResNet and Vision Transformer architectures show that Robust CurveMoE consistently improves clean, norm-specific, and Union accuracy over MSD and ERMC. In particular, it improves Union accuracy by 2.37 and 2.13 percentage points over the strongest baseline on CIFAR-100 and ImageNet-100, respectively. Extensive ablations further validate the effectiveness of partial updating, selective expertization, and robustness-constrained expert selection.