Research map / Adversarial robustness and security / Certified and empirical robustness
Error Correcting Output Codes Improve Probability Estimation and Adversarial Robustness of Deep Neural Networks
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 |
|---|---|---|---|
| replace softmax with sigmoid (logistic) decoding (ablation) | one-hot CNN · MNIST | CW accuracy: .540 -> .660 | Helped |
| replace one-hot code with 16-bit Hadamard code and tanh (ablation) | Logistic CNN · MNIST | PGD accuracy: .093 -> .421 | Helped |
| train an ensemble of small bit networks instead of one output layer (ablation) | Logistic CNN · MNIST | DAA accuracy: .684 -> .905 | Helped |
| combine Hadamard code with ensemble (TanhEns16 vs Tanh16) (ablation) | Tanh16 · MNIST | PGD accuracy: .421 -> .929 | Helped |
| increase code length from 16 to 32 bits (ablation) | TanhEns · MNIST | PGD accuracy: .929 -> .898 | No clear effect |
| increase code length from 16 to 32 bits (ablation) | TanhEns · CIFAR10 | PGD accuracy: .515 -> .574 | Helped |
| increase code length from 32 to 64 bits (ablation) | TanhEns · CIFAR10 | PGD accuracy: .574 -> .601 | Helped |
| use TanhEns16 instead of adversarially trained Madry model (ablation) | Madry · MNIST | PGD accuracy: .925 -> .929 | No clear effect |
| use TanhEns16 instead of adversarially trained Madry model (ablation) | Madry · CIFAR10 | PGD accuracy: .470 -> .515 | Helped |
| use TanhEns16 instead of softmax baseline (ablation) | Softmax · CIFAR10 | benign accuracy: .864 -> .888 | Helped |
Closest papers by meaning
- Improving Adversarial Robustness Requires Revisiting Misclassified Examples (ICLR 2020 · similarity 0.91)
- Certified Adversarial Robustness with Additive Noise (NeurIPS 2019 · similarity 0.91)
- Theoretically Principled Trade-off between Robustness and Accuracy (ICML 2019 · similarity 0.90)
- Are Adversarial Examples Created Equal? A Learnable Weighted Minimax Risk for Robustness under Non-uniform Attacks (AAAI 2021 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.90)
- A Direct Approach to Robust Deep Learning Using Adversarial Networks (ICLR 2019 · similarity 0.89)
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.