Research map / Adversarial robustness and security / Certified and empirical robustness
Ensemble-in-One: Ensemble Learning within Random Gated Networks for Enhanced Adversarial Robustness
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use adversarial feature distillation (ablation) | EIO training on CIFAR-10 · CIFAR-10 | Black-box adversarial accuracy: 0 -> 69.8 | Helped |
| Use adversarial feature distillation (ablation) | EIO training on CIFAR-10 · CIFAR-10 | Clean accuracy: 91.2 -> 84.7 | Hurt |
| Add random gating to the distillation-trained network (ablation) | EIO training on CIFAR-10 · CIFAR-10 | Clean accuracy: 84.7 -> 87.3 | Helped |
| Add mutual learning between paths (ablation) | EIO training on CIFAR-10 · CIFAR-10 | Black-box adversarial accuracy: 70.3 -> 75.4 | Helped |
| Increase the number of sampled paths per iteration (ablation) | EIO · CIFAR-10 | Black-box adversarial accuracy | Helped |
| Increase the number of sampled paths per iteration (ablation) | EIO · CIFAR-10 | Clean accuracy | Hurt |
| Fine-tune the extracted sub-model on clean examples (ablation) | EIO · CIFAR-10 | Clean accuracy | Helped |
| Fine-tune the extracted sub-model on clean examples (ablation) | EIO · CIFAR-10 | Black-box adversarial accuracy | No clear effect |
| Use EIO instead of standard training | ResNet-20 · CIFAR-100 | Black-box adversarial accuracy: 0.7 -> 51.7 | Helped |
| Use EIO instead of DVERGE | ResNet-18 · Tiny-ImageNet | Black-box adversarial accuracy: 13.1 -> 33.7 | Helped |
| Use EIO instead of AdvT | ResNet-20 · CIFAR-100 | Black-box adversarial accuracy: 49.7 -> 51.7 | Helped |
| Use EIO rather than DVERGE/8 for training | ResNet-20 · CIFAR-10 | Training time: 49.3h -> 10.9h | Helped |
Closest papers by meaning
- MORA: Improving Ensemble Robustness Evaluation with Model Reweighing Attack (NeurIPS 2022 · similarity 0.90)
- Improving Adversarial Robustness via Promoting Ensemble Diversity (ICML 2019 · similarity 0.89)
- Ensemble Adversarial Training: Attacks and Defenses (ICLR 2018 · similarity 0.89)
- AUTE: Peer-Alignment and Self-Unlearning Boost Adversarial Robustness for Training Ensemble Models (AAAI 2025 · similarity 0.87)
- On the Diversity of Adversarial Ensemble Learning (ICML 2025 · similarity 0.86)
- Two is Better than One: Efficient Ensemble Defense for Robust and Compact Models (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.