Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Margin Maximization Networks
Ziang Yan, Yiwen Guo, Changshui Zhang
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 |
|---|---|---|---|
| use MIN aggregation instead of AVG with LIN shrinkage (ablation) | AMM on MLP · MNIST | error rate (%): 2.26 -> 1.28 | Improved |
| add AVG+LIN regularizer to the MLP (ablation) | MLP baseline · MNIST | error rate (%): 1.79 -> 2.26 | Hurt |
| use EXP shrinkage instead of LIN under AVG (ablation) | AMM on MLP · MNIST | error rate (%): 2.26 -> 0.94 | Improved |
| use INV shrinkage instead of LIN under AVG (ablation) | AMM on MLP · MNIST | error rate (%): 2.26 -> 1.18 | Improved |
| use EXP shrinkage instead of LIN under MIN (ablation) | AMM on MLP · MNIST | error rate (%): 1.28 -> 0.90 | Improved |
| use MIN instead of AVG aggregation with EXP (ablation) | AMM on MLP · MNIST | error rate (%): 0.94 -> 0.90 | Improved |
| back-propagate through grad f (high order gradients) (ablation) | AMM on MLP · MNIST | error rate (%): 1.46 -> 0.90 | Improved |
| add AMM to MLP trained on 20k images vs vanilla on 60k | MLP · MNIST | error rate (%): 1.81 -> 1.54 | Improved |
| add AMM when 55k of 60k labels are random | MLP · MNIST noisy labels | error rate (%): above 40% -> below 10% | Improved |
| add AMM to LeNet | LeNet · MNIST | error rate (%): 0.87 -> 0.56 | Improved |
| add AMM to NIN | NIN · CIFAR-100 | error rate (%): 37.75 -> 34.49 | Improved |
| add AMM to ResNet-20 | ResNet-20 · CIFAR-100 | error rate (%): 33.20 -> 32.96 | No clear effect |
Closest papers by meaning
- Analyzing the Implicit Bias of Adversarial Training From a Generalized Margin Perspective (TPAMI 2025 · similarity 0.91)
- Recent Advances in Large Margin Learning (TPAMI 2021 · similarity 0.91)
- MMA Training: Direct Input Space Margin Maximization through Adversarial Training (ICLR 2020 · similarity 0.91)
- Adversarial Learning With Margin-Based Triplet Embedding Regularization (ICCV 2019 · similarity 0.90)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.87)
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.
