Research map / Optimization / Implicit bias and generalization
On the Convergence of Certified Robust Training with Interval Bound Propagation
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 |
|---|---|---|---|
| Train with IBP using sufficiently small perturbation radius and sufficiently large network width | Two-layer overparameterized ReLU network with logistic loss · Binary classification training set | training error | Helped |
| Use IBP training instead of standard training at the same width | Two-layer fully-connected ReLU network · MNIST (digits 2 and 5) | training error | Hurt |
| Increase perturbation radius (ablation) | IBP training · MNIST (digits 2 and 5) | training error | Hurt |
| Increase network width up to 80,000 while using ε=0.04 (ablation) | IBP training · MNIST (digits 2 and 5) | training error | No clear effect |
| Increase perturbation radius beyond the theorem's sufficient upper bound | IBP training · Binary classification training set | convergence to zero training error | Hurt |
| Set ε=0 (ablation) | IBP training, equivalent to standard training · Binary classification training set | training error | Helped |
Closest papers by meaning
- Understanding Certified Training with Interval Bound Propagation (ICLR 2024 · similarity 0.92)
- Towards Stable and Efficient Training of Verifiably Robust Neural Networks (ICLR 2020 · similarity 0.92)
- Robust Training of Neural Networks against Bias Field Perturbations (AAAI 2023 · similarity 0.91)
- Towards Evaluating and Training Verifiably Robust Neural Networks (CVPR 2021 · similarity 0.90)
- Expressive Losses for Verified Robustness via Convex Combinations (ICLR 2024 · similarity 0.84)
- Convex Formulations for Training Two-Layer ReLU Neural Networks (ICLR 2025 · similarity 0.84)
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.