Research map / Adversarial robustness and security / Black-box and transfer attacks
Error-Correcting Output Codes with Ensemble Diversity for Robust Learning in Neural Networks
Yang Song, Qiyu Kang, Wee Peng Tay
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Add ensemble diversity regularization (ablation) | ECNN · CIFAR-10 PGD attack | accuracy (%): 17.0 -> 41.2 | Improved |
| Add parameter sharing at both ends of the encoder (ablation) | ECNN^30_0.1 · CIFAR-10 PGD, ε=0.04 | accuracy (%): 51.1 -> 71.6 | Improved |
| Add parameter sharing at both ends of the encoder (ablation) | ECNN^30_0.1 · CIFAR-10 normal | accuracy (%): 79.2 -> 85.1 | Improved |
| Increase ensemble size from 10 to 30 binary classifiers (ablation) | ECNN^N_0.1 · CIFAR-10 PGD, ε=0.04 | accuracy (%): 41.2 -> 71.6 | Improved |
| Combine ECNN with adversarial training (ablation) | ECNN^30_0.1 · CIFAR-10 PGD, ε=0.03 | accuracy (%): 59.6 -> 69.4 | Improved |
| Use ECNN rather than TanhEns16 | MNIST · MNIST PGD, ε=0.3 | accuracy (%): 79.2 -> 88.4 | Improved |
| Use ECNN rather than TanhEns16 | CIFAR-10 · CIFAR-10 PGD, ε=0.04 | accuracy (%): 63.1 -> 71.6 | Improved |
| Use ECNN rather than ResNet20 | CIFAR-100 · CIFAR-100 normal | accuracy (%): 62.2 -> 61.3 | Hurt |
| Use ECNN rather than ResNet20 | CIFAR-100 · CIFAR-100 PGD, ε=0.04 | accuracy (%): 5.3 -> 36.3 | Improved |
| Use ECNN rather than ResNet20 | GTSRB · GTSRB OptProjTran | accuracy (%): 48.2 -> 79.6 | Improved |
Closest papers by meaning
- Bridging the Gap: Unifying the Training and Evaluation of Neural Network Binary Classifiers (NeurIPS 2022 · similarity 0.85)
- Error Correcting Output Codes Improve Probability Estimation and Adversarial Robustness of Deep Neural Networks (NeurIPS 2019 · similarity 0.85)
- A Simple Class-Agnostic Approach to Enhance Fair Adversarial Training (NeurIPS 2026 · similarity 0.85)
- Error Correction Output Codes for Robust Neural Networks against Weight-errors: A Neural Tangent Kernel Point of View (NeurIPS 2024 · similarity 0.84)
- Gradient Flow Provably Learns Robust Classifiers for Orthonormal GMMs (ICML 2025 · similarity 0.83)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.83)
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.
