Research map / Adversarial robustness and security / Certified and empirical robustness
Fundamental Tradeoffs in Distributionally Adversarial Training
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 an ℓ2 Wasserstein adversary (r=2) instead of other perturbation norms | Linear classifiers under the Gaussian mixture model · Gaussian mixture binary classification | standard and adversarial risk tradeoff | Helped |
| Increase the random-feature network width N | Random features model · Learning random quadratic functions | standard risk | Helped |
| Increase the random-feature network width N | Random features model · Learning random quadratic functions | adversarial risk | Helped |
| Increase the random-feature network width N | Random features model · Learning random quadratic functions | tradeoff between standard and adversarial risk | No clear effect |
| Increase adversary power ε | Linear regression models · Linear regression | range of Pareto-optimal estimators | Helped |
| Increase adversary power ε | Linear classifiers · Gaussian mixture binary classification | range of Pareto-optimal models | Helped |
| Increase feature correlation ρ | Linear regression models · Linear regression | standard and adversarial risk at λ=0 | Mixed |
Closest papers by meaning
- Adversarial Training and Robustness for Multiple Perturbations (NeurIPS 2019 · similarity 0.90)
- Explicit Tradeoffs between Adversarial and Natural Distributional Robustness (NeurIPS 2022 · similarity 0.90)
- Better Safe Than Sorry: Preventing Delusive Adversaries with Adversarial Training (NeurIPS 2021 · similarity 0.89)
- Combining Adversaries with Anti-adversaries in Training (AAAI 2023 · similarity 0.89)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- Adversaries With Incentives: A Strategic Alternative to Adversarial Robustness (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.