Research map / Learning theory and algorithms / Kernels and regression theory
The Price of Robustness: Stable Classifiers Need Overparameterization
Jonas von Berg, Adalbert Fono, Massimiliano Datres, Sohir Maskey, Gitta Kutyniok
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 |
|---|---|---|---|
| increase MLP width (interpolating networks) | 8-layer MLP · CIFAR-10 | class stability S(f) and normalized co-stability | Improved |
| increase CNN width | 4- and 8-layer CNN · CIFAR-10 | class stability S(f) | Improved |
| increase width with Heaviside (discontinuous) activations | 4- and 8-layer Heaviside MLP · MNIST | class stability S(f) | Improved |
| increase variance of Gaussian data so isoperimetric scale c grows (ablation) | 4-layer MLP, width 128-2048 · Gaussian toy data (linear task) | class stability | Hurt |
| use weight norms as predictor instead of stability measures | MLP width scaling · CIFAR-10 | tracking of test accuracy | No clear effect |
Closest papers by meaning
- Robustness of classifiers: from adversarial to random noise (NeurIPS 2016 · similarity 0.89)
- Robustness between the worst and average case (NeurIPS 2021 · similarity 0.88)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.87)
- Building a stable classifier with the inflated argmax (NeurIPS 2024 · similarity 0.87)
- Algorithmic Stability Unleashed: Generalization Bounds with Unbounded Losses (ICML 2024 · similarity 0.87)
- Certification-Enhanced Generalization Bounds (NeurIPS 2026 · 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.
